Location : Home > Resource > Paper > Theoretical Deduction
XU Duoqi | A Double-layered Mechanism of Fau​lt Tolerance in RegTech
2024-03-02 [author] XU Duoqi preview:

[author]XU Duoqi


A Double-layered Mechanism of Fault Tolerance in RegTech

XU Duoqi

Professor of Fudan University Law School

Abstract: The development of the digital economy has entered a phase of "deepened application, standardized development, and inclusive sharing", While Finlech and Reglech are still in an asynchronously evolutionary process. The innovation of RegTech is prone to failure, which has its own unique characteristics and deep-rooted causes, thus it is necessary to strengthen the double-layered mechanism of fault tolerance in terms of RegTech and FinTech. Faced with trilemma in objectives of regulatory authorities, third-party technology providers, and regulated institutions, the Reglech and ComTech is facing the predicament of unsuccessful connection, thus it is necessary to make a shift towards a concept of tolerant regulation. Tolerant regulation does not mean "non-regulation" or "campaign-oriented regulation", instead it is a kind of regulation that provides a relative margin of error according to legal procedure, encourages innovation, and promotes the regulation capable of realizing sustainable development, with a focus on the overall social interests and stable expectations in normal regulatory practices. In the double-layered mechanism of fault tolerance, the Central Financial Commission and the Financial Work Committee serve as the tolerance body in the upper layer, while the object of tolerance and exemption is the financial regulators who take responsibility to act and are brave enough to innovate. The content of tolerance should shift from purely developing regional experimental fields to comprehensively promoting the simultaneous development of RegTech innovation and risk prevention. At the same time, the priority of procedures and the precision of accountability should be emphasized.

Keywords: Fintech; RegTech; ComTech; Double-layered Mechanism of Fault Tolerance; Inclusive Regulation

1. Background and Issues

While financial regulation is one of the policy areas where AI is widely used, the emergence and development of Regulatory Technology(RegTech) is significantly lagging behind that of Financial Technology (FinTech), and remains one of the newer generations in the transformation of technology-enabled industries. The UK Financial Conduct Authority (FCA) believes that RegTech aims to facilitate the implementation of regulatory requirements more efficiently through technological means, and is a new technology developed to overcome the challenges of financial services regulation. In essence, financial technology is changing rapidly, new products, new channels and new forms of business continue to emerge, financial risks have more prominent complexity, contagiousness, externality and systemic characteristics, the traditional regulatory concepts and regulatory tools are unable to solve the problem of adaptive regulation. Fintech mainly consists of two types: one is the use of technological innovation by fintech companies to provide technical solutions to finance; the other is the use of technology by traditional financial institutions to improve their own products and services, enhance efficiency and reduce costs. However, while fintech contributes to improving the quality and efficiency of financial services, its technological application and business transformation may also increase the vulnerability of traditional financial institutions and exacerbate the original risks; technological innovation may also lead to more frequent regulatory failures, resulting in regulatory loopholes or deriving regulatory arbitrage and other phenomena. Regulatory authorities have found through their own practice that they need to develop (often entrusted to third-party organizations) or open up innovative technological tools suitable for regulation, otherwise it is difficult to keep pace with the rapid development of financial technology.

In retrospect, from the regulation that accommodated the rapid growth of Internet finance to the regulation in motion, it is like the two ends of a pendulum, swinging suddenly from one end to the other without any transition or stay in between, with far-reaching impacts on the market. Over-regulation will lead to choking and the phenomenon of bad money driving out good money. How to adopt a normalized and inclusive regulatory approach that does not go to extremes has always been a practical issue that cannot be ignored. In this respect, it is crucial to build a set of fintech fault-tolerance mechanism that can both prevent risks and reasonably delimit the space for free innovation. The so-called fault-tolerance mechanism refers to a dynamic interrelationship that accommodates and tolerates mistakes or errors. In 2018, the State Council's Opinions on Comprehensively Strengthening Basic Scientific Research explicitly proposed to "establish a fault-tolerance mechanism that encourages innovation and tolerates failures, and encourage researchers to boldly explore and challenge the unknown." This should also apply to the field of regulatory science and technology. In August 2019, the People's Bank of China (hereinafter referred to as the Central Bank) issued the first round of financial technology development planning - "Financial Technology (Fintech) Development Planning (2019-2021)", which puts forward the guiding ideology, basic principles, development goals and key tasks of the next three years of work in financial technology, In January 2022, the Central Bank issued the Fintech Development Plan (2022-2025), aiming to promote fintech from "building pillars and beams" to a new stage of "accumulating momentum", and to expand the scope of innovative regulatory pilots. Both rounds of financial technology development planning reflects the central bank's attitude of inclusive regulation, and here the so-called inclusive regulation is obviously only the fault tolerance of the underlying financial technology application itself, rather than the fault tolerance at the level of regulatory science and technology, which may still lead to the tendency of the regulatory authorities to pass the buck to each other. In the author's view, the current dilemma facing financial regulation is no longer limited to the problem of fault-tolerant regulation of fintech innovation pilots, but also includes the problem of fault-tolerant mechanism of regulatory technology itself. The connotation of "error" here involves two levels: one is the fault tolerance between the regulatory authority and the regulated organization, i.e., fintech trial or sandbox regulation; the other is the unavoidable mistakes made by the regulatory authority in regulating financial institutions in cooperation with the third-party technology enterprises, even though the regulatory authority is fulfilling its duty, and the unavoidable mistakes made when carrying out the pilot regulatory innovation in finance. 

In this regard, the author tries to summarize this system with hierarchical linkage with a two-tier fault-tolerance mechanism: on the one hand, conditions should be created for FinTech as a regulatory target to make it trial and error and develop within a controllable scope of influence; on the other hand, RegTech as a regulatory tool should be given a certain degree of fault-tolerance space, so as to enable iteration and innovation in the course of sufficient feedback and correction. While much attention has been paid to fault-tolerance mechanisms for fintech, such as "sandbox regulation", there is still a lack of research on fault-tolerance at the level of innovation in regulatory technology. How should regulators deal with the contradiction between "stability of financial legislation and flexibility of its application", "trial and error of innovation regulation and regulatory risk control", and "sport regulation and inclusive regulation"? Once the continuous breakthroughs in technological capabilities bring unpredictable and huge risks, can regulators change their regulatory concepts, work hand in hand with technology to reduce compliance costs and achieve effective regulation, and make regulatory technology an important digital scenario and a "testing ground" for the application of digital technological innovations? A positive answer to these questions is a more fundamental and realistic issue of contemporary significance than how to design a single financial innovation and its risk prevention measures.

2. The inevitability of strengthening the two-tier fault-tolerance mechanism for regulatory technology

Regulatory technology, as a new method of regulation that carries technological innovation, has emerged, advocating the integration of law and technology to jointly improve the government's financial regulatory capacity, through the introduction of the technological power of third-party technology enterprises, together with the regulatory authorities to regulate the innovation of financial science and technology, changing the insufficient status of the government's sole reliance on the rules of law to regulate in the field of financial regulation, and realizing a breakthrough and transcendence of the dichotomy between the government and the market in the traditional regulation. In order to reflect the common intrinsic needs of the government and the market, the establishment and implementation of fault-tolerance mechanism is indispensable for this new regulatory concept and regulatory approach. As people analyze the reasons for the failure of regulatory science and technology, they tend to limit their vision to the failure of the regulatory authority or the market operation itself and its reasons, therefore, building a two-tier fault-tolerance mechanism analytical framework of heterogeneous interaction is not only a call for theoretical innovation but also an inevitable response to the pain points of practice.

2.1 Why RegTech is prone to failure

Only by fully understanding the types and causes of failures in regulatory science and technology, and recognizing which types of failures are tolerable and which are not, can the inherent value balance between "defense and enterprise" be sought. In addition to the financial constraints of the public sector, innovation in regulatory science and technology faces a variety of constraints that are different from those of scientific and technological innovation.

First, there is a lack of intrinsic motivation for regulatory competition. Regardless of whether it is unified regulation, bimodal regulation or sectoral regulation, financial regulatory authorities usually have their own responsibilities and occupy a monopoly position in their respective areas of competence. Even with the umbrella regulatory features of the U.S. financial regulatory system, the Office of the Comptroller of the Currency, the Federal Deposit Insurance Corporation and the Federal Reserve and other regulatory authorities also exist more or less competition, the U.S. Securities and Exchange Commission (SEC) and the Commodity Futures Trading Commission (CFTC) is no lack of regulatory competition history. The fragmented and conflicting regulatory system in the United States has been subject to much criticism, and while proponents argue that it is still holding the line under the banner of "regulatory competition," opponents criticize the competition between agencies for attracting the private sector or investment, which has ultimately become a "race to the bottom" game, a tendency that is dynamically reflected in the regulatory sandboxes and innovation hubs of the regulatory agencies, and that this kind of amplified competition can lead to bottomless and borderless regulation. In other words, public officials will not be frustrated by a lack of innovation, but only a strong desire to achieve public goals can stimulate regulatory innovation.

Second, it is an external manifestation of regulatory risk avoidance. Financial regulators also face the practical dilemma - the list of public targets is long and varied, including consumer protection, currency stability, financial stability, market efficiency, crime prevention, and so on, to name but a few. Each of these objectives is more difficult to promote than most private sector innovations, and regulators are often reluctant to exercise their regulatory responsibilities or even pass the buck due to conflicting regulatory objectives, difficult trade-offs or unclear responsibilities. Needless to say, the rotation of public sector policymakers every few years, which is common in all countries, makes it difficult to obtain effective long-term incentives for projects that do not achieve significant results in the short term, and the annual reporting system, which makes regulators less incentivized to consider the longer-term consequences of a law or a policy because of their eagerness to meet the assessment targets, short-termism restricts to a greater extent the incentives to innovate of the regulators.

Third, regulatory rule-obsession. This is essentially a Kafka style bureaucracy with risk aversion and adherence to conventions. If the rules and procedures themselves are viewed as goals, in order to avoid the risk of uncertainty, innovation will inevitably encounter huge obstacles, and the tendency to inaction will increase significantly, leading to more regulatory traps and innovation dilemmas. In France, the Prudential Regulation Authority (ACPR) has established an "internal entrepreneurship" project aimed at encouraging employees to suggest or lead innovative projects, improve ACPR tools and application processes; Le Lab is responsible for the design of selected projects and provides dedicated sponsors, external coaches, and IT support. The European Central Bank has established a super technology virtual laboratory to conduct super technology experiments. These constantly emerging innovation incubation projects all come from inclusive policy support provided by the government. There are few professional fintech incubator projects similar to LabCFTC in the United States, and the Commodity Futures Trading Commission helps the private sector develop fintech rather than promoting fintech experimentation. The Federal Deposit Insurance Corporation is also seeking to collaborate with the financial industry to develop new regulatory compliance reporting tools, and is beginning to experiment with trial and error and fault-tolerant fintech innovation projects. Although adopting such financial technology fault tolerance strategies is not a panacea, it can to some extent help regulatory authorities break free from the institutional culture of being obsessed with rules and risk averse, and also make people with innovative genes attractive to fault tolerance mechanisms and prioritize cooperation with regulatory authorities.

So, if regulatory technology fails, leading to inadequate regulation, regulatory capture, and other phenomena, how can we tolerate mistakes? Can a cooperative double-layer fault-tolerant mechanism that balances internal and external factors and encourages compatibility solve the problem of regulatory technology project failures?

2.2 The necessity of implementing a dual fault tolerance mechanism by regulatory authorities

The fault-tolerant mechanism of regulatory technology pursues a dynamic balance between efficiency and fairness, which can provide necessary development time and trial and error space for effective cooperation between regulatory authorities, regulated institutions, and third-party technology enterprises. The policy guidance in financial regulation is incentive oriented, although it is economic, it implies a political tone. The biggest risk in regulatory technology is the risk of resource allocation. If precious public resources are used for innovative projects, the responsible parties may worry about taking responsibility for wasting resources and instead allocate them for other purposes. Due to the scarcity of resources, it is urgent for regulatory authorities to spend greater costs to promote innovation, leading to a more complex situation of innovation failure or loss of control for regulatory technology. It can be seen that the fault-tolerant mechanism here is not limited to the fault-tolerant space of financial technology innovation between regulators and regulated entities, but also lies in the implementation of upper level fault-tolerant mechanisms through regulatory innovation by regulators.

Firstly, regulatory technology should evolve in sync with fintech, otherwise it is highly likely to halt fintech innovation. Compared with developed countries, China's Internet finance started late. In order to support innovation and promote the development of new formats, the regulatory authorities adopted an inclusive regulatory approach. Inclusive regulation is a "new governance" regulatory paradigm derived from embedding the concept of "financial inclusion" into the financial regulatory framework system. Since "Internet finance development and supervision" first entered one of the 19 research topics in the financial field deployed by the State Council in April 2013, the term "Internet finance" has only begun to enter the vision of regulators. Against the backdrop of the "mass innovation" promotion policy, China's financial repression has led to high financing costs and large financing demand for SMEs, while the wealth of Chinese households is rising, and diversified investment channels are obviously insufficient. Under the attitude of regulatory inclusiveness, P2P and other Internet lending platforms continue to alienate and grow savagely, from a dozen to more than 5000. Since 2015, problems such as large-scale platforms "running away", false lending and fraud, illegal self financing and other issues have broken out in succession, illegal absorption of public deposits, fund-raising fraud and other incidents have occurred in succession, which directly led to the launch of a special rectification action for Internet financial risks in October 2016, that is, sports supervision in this field. Subsequently, in December 2018, the "Opinions on Doing a Good Job in Classifying Disposal and Risk Prevention of Online Loan Institutions" (Rectification Office Letter [2018] No. 173) was issued, requiring online loan institutions to exit or transform based on their own risk situation and business situation; In 2019, the Guiding Opinions on Pilot Transformation of Online Loan Information Intermediaries into Small Loan Companies (Rectification Letter [2019] No. 83) were issued, guiding some eligible online loan institutions to transform into small loan companies. Until November 2020, P2P lending institutions in actual operation nationwide were completely shut down. The return to zero of P2P is not an isolated case. The Internet deposits, many cash loans, digital currencies and even various financial technology products were stopped in the later period. In essence, the regulatory technology could not keep up with the pace of financial technology, which made it difficult to distinguish between genuine and fake financial innovation. The only way to prevent further damage to the market caused by fake financial innovation was to stop.

The second is that the probability of failure in regulatory technology and fintech is not on the same level. Successful regulation is often difficult to detect, but failure can trigger a huge financial crisis. At the same time, the probability of successful regulatory innovation is disproportionate to the continuous refurbishment of destructive methods. On the one hand, from the perspective of the relationship between offense and defense, the innovation of regulated entities belongs to offensive innovation, and as long as there is one breakthrough in one hundred offensive innovations, it can be considered successful; Regulatory innovation belongs to defensive innovation, and as long as it is breached once, it means failure. Therefore, the probability and degree of failure of the two cannot be compared. On the other hand, regulators are limited by regulatory funds and regulatory capabilities, which can lead to regulatory competition and regulatory capture, resulting in the failure of regulatory technology; If outsourced to the private sector, there is a high possibility of vicious competition and regulatory arbitrage failure. Even if outsourced to academic institutions rather than non-profit enterprises, although concerns about regulatory arbitrage are reduced, the practicality, transferability, and degree of compliance with expectations of academic institution results still need to be tested. In addition, creating new technologies on top of traditional financial regulatory systems may pose unpredictable risks, and even small interface interoperability barriers between old and new systems may hinder technological innovation.

Thirdly, regulatory technological innovation often lacks motivation and requires more fault tolerance mechanisms. One is talent mismatch. Regulatory technology requires regulatory agencies to have sufficient financial, legal, and technical personnel, but financial regulatory agencies mainly have public law lawyers, economists, and accountants, rather than software engineers or data scientists. The lack of talent has led to internal difficulties in developing auxiliary technology solutions, and the talent support capacity for regulating technological innovation is seriously insufficient. Then, regulatory thinking bias. The inherent rule obsession and risk aversion deficiencies of regulatory authorities may lead to resource constraints, and even if suitable third-party technology development companies are selected, the quality of technology depends on the regulatory authority's budget and the ability to monitor programming and technology development of partners or suppliers. If the effectiveness of regulatory technology is based on the fairness, legality, and credibility of widely used machine learning algorithms, then these "black box" algorithms face an inherent dilemma. Even if the algorithm developers know the input data and algorithm results, they may not fully understand the process of algorithm output results, making it difficult to determine causal relationships, As a result, it cannot solve the problem of ensuring that regulatory technology meets people's expectations of democratic accountability. Finally, bias in automated decision-making. Regulatory technology pays special attention to the application of machine learning. Once machine learning algorithms are trained, the users of the algorithms must give them full trust. In the field of regulatory technology, algorithms evaluate identity and reputation through classification analysis, but they may also create opportunities for discrimination, normalization, and manipulation due to a lack of transparency, accountability mechanisms, detection systems, and due process constraints. These practical dilemmas largely reflect the impact of regulatory technology's substantive rationality and legal perspective on financial regulatory norms. The fundamental solution lies in not following traditional financial regulatory ideas, but considering the special circumstances of the actors, adjusting and correcting identified illegal behaviors for different actors, and achieving the unity of general normative adjustment and individualized consideration of the law.

Further, for macroprudential purposes, regulatory science and technology experiments are generally slow to start and move forward. In contrast to the financial solvency regulation and liquidity risk regulation of micro prudential regulation, macroprudential regulation usually focuses on higher value tasks. Potential conflicts in the consideration of individual institutions' risk management strategies can lead to systemic risk issues, and because of the higher requirements for assessing systemic risk, relevant regulatory innovation is more scarce. Of course, this is not to say that there are no experiments with macroprudential techniques. For example, the Federal Reserve and the Bank of England are using natural language processing to parse large volumes of documents to predict trends, and the Bank of Italy is experimenting with machine learning to analyze "the popularity of real estate advertisements" to predict house price movements and inflation trends. European banks are even exploring the use of quantitative analysis of market sentiment to enhance financial risk forecasting. A more general regulatory need is a dashboard-like "visual" regulatory forecast chart, which is useful for monitoring "the combination of micro-regulatory purposes of different developments that require and may conceal unseen risks," making the long-term control of financial stability more precise and clear. At the same time, financial regulators are also constantly trying to use the power of science and technology to improve the real-time performance of financial reporting and risk disclosure, and transform the delayed monitoring after the occurrence of damage into the proactive prevention of early intervention.

It can be seen from the above that in addition to the fault-tolerant mechanism required for fintech innovation itself, regulatory technology also faces many difficulties and needs a fault-tolerant mechanism. Not only is the probability of potential failure of regulatory technology far greater than that of financial technology, but because of the numerous public goals pursued, including financial stability, consumer rights protection, market efficiency and fairness, anti-money laundering and anti-fraud, it is difficult to choose between various public goals, and regulatory technological innovation is like "singing on stilts", and uncertainty is difficult to measure. While attaching great importance to the probability of failure, we should not give up the innovative application of regulatory technology for fear of choking. On the contrary, the reform exploration cannot "only allow success and not failure", and should promote a two-tier fault tolerance mechanism under the governance concept of "focusing on encouraging innovation and tolerating mistakes". In fact, there have been more extensive flexible inclusive regulatory attempts than the "regulatory sandbox", allowing the regulated to "break the law" and "trial and error", and the regulators and the regulated to achieve consultation and communication, interactive trial and error, and innovative regulation in the discussion and game. The financial supervision methods implemented in China, including interviews, suggestions, guidance, etc., have long gone beyond the "sandbox" and have gone even further. For regulatory technology, the fault-tolerant mechanism should be an integral part of inclusive regulation, an adaptive measure to deal with the conflict of diversified subject objectives of regulatory technology, and an adequate institutional environment for good laws and good governance for financial regulatory reformers to dare to reform and innovators to innovate.

3. Triple Goal Conflict: Misalignment of Regulatory Technology and Compliance Technology

In the practice of law enforcement duties, there are types of "good faith violations" that allow for fault tolerance in cases where the purpose is legitimate and exceeds authority, the purpose is legitimate and reduces procedures, and the purpose is legitimate and violates substantive law. However, in financial regulatory technology, the fault tolerance mechanism faces the challenge of multiple purposes. As is well known, regulatory technology attempts to transcend the binary opposition between government and market and achieve the regulatory goal of taming financial technology power, stemming from its own technological background and driving mechanism. However, the goals of financial regulatory policies exhibit diversified characteristics. If the policy supporting measures are difficult to keep up with the development of practice, and the goals pursued by regulators, regulated parties, and cooperating technology third parties are also different, sufficient trial and error space must be left, otherwise it is easy to cause centrifugal connection failure between regulatory technology and compliance technology among the three parties.

3.1 Triple target conflict

The most typical implementation method of regulatory technology is a collaborative environment composed of regulatory authorities and industry sandboxes under technical support, while the mechanism for third-party enterprises to provide technical support is authorized by regulatory technology and compliance technology to comply with regulatory and compliance reporting requirements and promote supervisory obligations. The joint supervision achieved by regulatory authorities and third-party enterprises based on technology can help achieve diverse policy goals, including market fairness and efficiency, financial stability, consumer and investor protection, law enforcement efficiency, and technological innovation. The collaborative regulatory approach supported by technology is clearly superior to traditional command and control methods. However, the DingBergen rule suggests that one tool is most efficient in achieving one policy objective, and if one tool is attempted to achieve more than one policy objective, efficiency will be reduced due to conflicting objectives, and even deviate from the objectives, resulting in a more imbalanced state. In the process of implementing regulatory technology, there is a challenge of divergent goals among the three main entities.

Firstly, regulatory authorities should pay more attention to policy objectives such as "public interest" and grant them the power of "fault tolerance", but they should not abuse this power. Nornett and Selznick once defined the contribution of responsive law as "promoting the achievement of public goals and casting the spirit of self correction into the process of government management". Regulatory authorities are different from private sectors that pursue maximum benefits. They often focus on providing public goods that are conducive to consumer protection and financial stability, while also possessing strong power to judge what financial innovation is. Especially when addressing social issues of strong public concern, they have a stronger political appeal. The necessity of regulating technology is unquestionable, but financial regulatory laws naturally lag behind the development of financial market practices, making it easy for "good faith illegal" behavior to occur. It is urgent to provide certain room for error tolerance and continuous trial and error. At the same time, the process of regulating the operation of technology itself is a process of power implementation, and the collision between individual adjustments and universal constraints of the law is more intense, leading to the inability of actors to perform their duties legally. In other words, regulatory authorities, in order to catch up with the development speed of financial technology, usually cooperate with third-party technology enterprises that focus on monitoring and analyzing the application of technology to generate real-time and insightful risk indicators, which are used to support the dynamic application of regulatory rule formulation, execution, and decision-making throughout the entire process, thereby achieving the regulation of financial fraud, protecting the rights and interests of financial consumers, and maintaining the stability of the digital driven financial system The diverse public policy objectives include ensuring financial security and guarding the bottom line of systemic risks, identifying and combating financial crimes, and providing biased protection for vulnerable groups. It even implies another important goal of regulatory technology, which is to promptly mitigate the adverse consequences of damage events after they occur. It can be seen that regulatory authorities have great power in promoting regulatory technology. In order to ensure the effective implementation of a dual fault tolerance mechanism in the event of target conflicts, they must not only limit the generalization of fault tolerance mechanisms caused by their power, but also prevent the virtualization of fault tolerance mechanisms caused by being held accountable.

Secondly, third-party technology enterprises aim to pursue the "cost-benefit" goals based on compliance innovation, and are in the middle of a dual fault tolerance mechanism. Due to the characteristics of cutting-edge, continuity, professionalism, and exploratory nature of innovation activities, third-party technology enterprises play an extremely important role in strengthening the supervision of scientific and technological innovation and coordinating the promotion of innovation fault-tolerant work. It may serve regulatory authorities and help them establish compliance enforcement models with legal management obligations as the core. For example, Swiss regulatory technology company Apiax has a relatively advanced regulatory technology system that can convert subjective and humanistic regulatory regulations into digital codes; It may also assist regulatory authorities in providing compliance technology incentives with responsibility reduction as the core for enterprises, and may also assist regulatory authorities or financial institutions in achieving compliance technology guidance with self-discipline as the core, such as using digital rules for regulatory management, conducting compliance behavior in cross-border financial activities, intelligent investment consulting, and other aspects. Strictly speaking, unlike the application of regulatory technology (RegTech) on the regulatory side, CompTech, located on the compliance side, focuses on helping regulated institutions improve their risk management capabilities and provide compliance capabilities, thereby serving as a bridge between regulatory authorities and regulated institutions. For example, TRAction in Australia will automate the processing of regulatory reports, achieve automatic conversion and supplementary compliance verification, and generate reports that can be automated for submission. It is worth mentioning that technological innovation in supporting financial services will not remain unchanged in operation, but will inevitably continue to develop after the technology is released. This brings about two consequences. On the one hand, third-party development suppliers only apply rules to the initial goals of the technology, lacking the motivation to intentionally provide sufficient space for public participation with social public interests. To some extent, regulatory authorities may trust large technology enterprises more, bringing greater trust costs to small and medium-sized technology enterprises. As observed by Mulligan and Bamberg, the initial discussion process "ignored the actions of regulatory authorities when they handed over the design and production of regulatory technology to standard setting agencies, engineers, designers, and project managers". In this way, there may be cognitive and regulatory mismatches between regulatory authorities and third-party technology companies. On the other hand, based on different interests, third-party technology companies may claim that the technology they develop is proprietary and cannot disclose too many trade secrets during the process of regulating the continuous development of technology. This poses challenges to technology transparency and public participation. When regulators and consumers come into contact with software code or other forms of complex technology, they may not only pass on costs to financial institutions with strong compliance needs, but may even have chain cost sharing consequences for their customers. Therefore, it is necessary to strengthen the indirect supervision of financial institutions and prevent technology-based enterprises involved in credit creation from artificially adjusting risk control model parameters to improve credit availability. Given the complexity of technology, public scrutiny and supervision will become almost impossible. As a result, there will inevitably be varying degrees of fragmentation between regulatory technology and compliance technology. It is crucial to establish and improve the synchronous evolution of innovative integrity systems and fault tolerance mechanisms.

Thirdly, regulatory authorities focus on compliance goals triggered by incentive and constraint conditions in the process of achieving policy objectives. Regulated individuals have a desire and ability to continuously drive innovation, and hope to maintain communication with regulators. With a cooperative attitude and compliant performance, they strive to gain the attention and trust of regulators, thereby providing an equal treatment, information transparency, and legal participation business environment for their innovative behavior. Due to the low understanding and weak implementation of regulatory policies by individual financial institutions, there may be deviations between their behavior and policies, which cannot meet the requirements. Regulatory technology can provide assistance in reducing compliance burdens while increasing the frequency of regulatory reporting, or by developing technical tools to implement regulatory policies. At present, financial institutions use privacy computing to integrate information such as business and judicial blacklists, revocation and cancellation, abnormal transactions, and dishonesty through joint modeling, establish fraud monitoring models, and integrate data from multiple parties to enhance fraud identification capabilities. However, no matter how compliance technology is promoted, there are still many illegal and non compliant phenomena due to different goals. With the continuous improvement of regulatory strength and level in recent times, some financial institutions fabricate and fabricate compliance data to meet compliance standards, resulting in distorted compliance data and subsequent judgment and decision-making deviations. According to statistics, the "strong supervision" trend in China's financial industry will continue in 2022, with regulatory authorities issuing consecutive fines. 285 fines are related to data quality, accounting for 61% of the total fines, ranking first in the list of punishment reasons, an increase of 132% compared to the same period last year. In addition, the number of penalties for failure to submit and comply with data regulations was 103 and 45, respectively, ranking second and third in terms of punishment reasons. Due to the fact that financial institutions are important market entities, negative labels imposed by regulatory and administrative penalties may lead to risks of being "too big to fail" and "too many links to fail", resulting in a "ripple like" social negative effect, causing employee unemployment, damage to the operations of partners and financing parties, and loss of industry reputation. This has accelerated the reshaping of the safety awareness of regulated institutions, which is no longer limited to passive compliance. Financial institutions are increasing their digital native security capabilities to ensure business security and stable operations through technology.

The conflict of ternary objectives can easily lead to the failure of the effectiveness of regulatory technology implementation, and the conflict between the public interest goals of regulatory authorities and the efficiency goals of third-party technology enterprises may not always lead to the smooth realization of the regulatory results they develop; There are also subtle differences in understanding between regulatory compliance goals and regulatory agencies, and there are significant differences in the driving forces behind capturing the potential meaning of the original version of regulatory policies, making it inevitable to avoid adaptation misalignment. Moreover, through technological integration, the fate of regulatory authorities and regulated institutions is intertwined. In addition to investing in their own technological operation systems, regulators must also supervise any third-party technology companies that provide supporting technological solutions, which contradicts the latter's right to advocate intellectual property protection or the need to interpret algorithms. As Professor Ranchodas from the Tilburg School of Law in the Netherlands once said, "Tolerance is a human characteristic. However, with the digitization of government services and the automation of government decision-making, the past empathy with the administrative state is disappearing." In other words, super technologies lacking error tolerance mechanisms may suffer greater failures, only on the basis of not breaking the law or "breaking the law in good faith". Only by adhering to a more tolerant regulatory philosophy can regulatory technology and compliance technology be cleverly combined to the greatest extent possible.

3.2 Dual connection failure

In addition to the universal confusion brought by technology to multiple entities, there is also the problem of improper connection between regulation and compliance. The fault-tolerant mechanism means allowing for the existence of "errors", and regulatory technological innovation that is completely free from conflicts and "errors" does not exist in reality. Under the conflict of the aforementioned three objectives, the problems that fault-tolerant mechanisms may face may include: how to promote regulatory technological innovation when regulatory powers do not act; how should the law protect the rights of consumers when technology treats them differently in the process of embedding regulation; should consumers trust automatic decision-making from black boxes, or should they trust human regulatory agencies working with black boxes; if regulatory authorities do not invest enough in maintaining the smooth operation of technological systems and ensuring consumer information security, how should they respond and so on. The solution to the aforementioned problem should not only be based on the explanation of the conflict relationship between the three objectives, but also seek the optimal solution for correcting the application of regulatory technology for financial consumers in the context of the failure of the connection between regulatory technology and compliance technology. Due to the different main objectives of the three parties mentioned above, there is often a mismatch or mismatch between technology and business, leading to the failure of the dual connection between regulatory technology and compliance technology.

On the one hand, there is an improper connection between the regulatory authorities and the regulated institutions in achieving their goals. One is the conflict between regulatory technology and compliance technology goals, which can easily lead to the failure of regulatory goals. Compliance does not necessarily mean the disappearance of all potential violations, and certain penalties are unavoidable. Therefore, in a sense, it is only seen as a cost, and the fault-tolerant mechanism is seen as a conditional incentive, balancing costs and benefits repeatedly to achieve the goal of relatively equal treatment. The regulatory authorities in Nepal use the method of collecting geographic data on the distribution of financial outlets to achieve the public goal of improving the accessibility of financial services to financial consumers. Financial institutions, as opposed to regulators, place greater emphasis on achieving compliance in credit loans. For example, banks collaborate with logistics companies to grasp the dynamic freight routes of truck drivers, and provide prepaid credit to truck drivers while controlling goods at any time. This greatly satisfies the accessibility of customer financial services while reducing non-performing loan ratios. Obviously, both regulators and regulated individuals collect geographic information, but their usage scenarios and implementation purposes are completely different, even going against each other, which can easily lead to non-compliance in terms of data collection purposes and other aspects. The second is that the direct purpose of regulators applying compliance technology is to meet the compliance needs of financial institutions, and to protect the rights of financial consumers in the absence of a unified information center for data integration. At the end of 2020, the chatbot BOB was launched in the Philippines, allowing financial consumers to file complaints through applications or mobile SMS services. The chatbot will accept complaints in different languages and assign case numbers or simple and complex sorting according to specific situations for processing. Complex complaint requests will be automatically upgraded and assigned to human services. But the problem is that in addition to complaints from chatbots, financial institutions will also face complaints from other sources such as voice calls and emails, both to the headquarters and from different regions. These complaints from different sources and entities are diverse and diverse, and there is no structured unified standard. If there is a lack of a processor similar to a central database, to assist management personnel of consumer protection agencies in summarizing and reviewing the analysis of chatbots, and to analyze and streamline their internal logic, regulatory policy goals related to consumer protection are easily overlooked due to being superficial. It can be seen that regulatory agencies may provide a large amount of scattered and redundant data for the purpose of compliance, but lack crucial, global, and collaborative data to connect different regulated entities. Finally, any open real-time compliance report based on machine learning may lead to overly complex report content or insufficient information. On August 22, 2022, the "Internal Control Management Measures for Wealth Management Companies" (Order No. 4 of the China Banking and Insurance Regulatory Commission in 2022) proposed the establishment of a Chief Compliance Officer system, which relies on a long-term combination mechanism of internal compliance and external supervision for its successful implementation. However, regulatory authorities, based on the purpose of prudent supervision, may be misled by redundant information or missing important information if they overly rely on natural language processing technology to review reports, and miss critical prevention and control opportunities that affect systemic financial risks. To some extent, the success, failure, and overall reputation of regulatory authorities depend on the behavior of the regulated institutions, and as technology integration develops, the fates of the two will become increasingly intertwined. In essence, the regulatory technology behavior of regulators is precisely to pursue the needs of the aforementioned public interest. The integration of technology and regulation improves the quality and efficiency of regulation, while not compromising the legitimate rights and interests of the regulated during the implementation process. Further protection of the regulated's desire and ability to promote innovation is necessary to form a corresponding regulatory technology fault tolerance space.

On the other hand, there is a mismatch in the goal alignment between regulatory authorities and third-party technology enterprises. Third party technology companies tend to focus on their own "cost efficiency" goals rather than "fair distribution of regulatory costs and benefits". Although they continuously improve data quality and launch innovative products to serve regulatory technology in the process of data collection, storage, analysis, and processing, they lack welfare management that pursues public interest goals in the field of auxiliary technology, and the regulatory technology they develop may deviate from regulatory goals. One is the phenomenon of neglecting the safety bottom line in the process of achieving regulatory goals. Third party technology companies should have made efforts to ensure that their readable versions "capture all potential connotations of the original version as much as possible without ambiguity," but their focus often leans towards broader application technologies, such as natural language processing (NLP), seeking patterns to find relevance in data, to "sort out a series of amazing materials, quickly find, summarize, classify, and present relevant information for further review." To achieve maximum compliance and efficiency. This can easily lead to third-party institutions ignoring the "public interest" goals upheld by regulatory authorities and turning to pursuing their own interests, thereby stimulating their motivation to break through the regulatory safety bottom line and threatening the security of regulatory technology by relying on their data or technological advantages. Secondly, regulatory authorities may abandon their firm commitment to public services in order to improve regulatory timeliness, while third-party technology companies may become more technical auxiliary tools in order to achieve accurate and accurate compliance goals. Regulatory authorities are exploring how to convert regulatory texts into machine-readable data, allowing computers to automatically read regulations without human guidance. The widespread use of technology may reduce regulatory empathy and impact their regulatory culture in the process of excessive reliance on technological tools. As regulatory technology work is increasingly delegated to third-party technology companies, regulations read and executed by computers will meet compliance reporting requirements more quickly. Even with increased compliance requirements, fewer employees can still meet the certainty requirements of compliance with the help of technology. Due to the increasing reliance on technology, regulatory authorities may also view their work as more technical, missing out on the risks that should be strictly regulated, or allocating irrelevant risk events based on technical indicators, while neglecting the creation and maintenance of a public service culture. Thirdly, regulatory authorities with limited resources tend to focus more on the quality and efficiency of regulation, while third-party technology companies are more concerned about achieving regulatory compliance goals. Regulatory authorities are interested in applications supported by APIs, cloud computing, and distributed ledger technology because APIs allow different types of software to communicate with each other, thereby improving interoperability; Cloud computing can provide protective protection for data storage - if a server in the network fails, as long as other servers in the cloud can compensate, the data will continue to be available; distributed ledger technology allows data to be stored in multiple locations, creating gaps (distributed ledger is essentially a database composed of multiple computers or corresponding "nodes", whose integrity is guaranteed by some form of consensus mechanism concatenated nodes), and the power to determine its innovation and legitimacy is in the hands of regulators. Regulatory authorities assume that, when all financial transaction records are stored in a distributed ledger, it is possible to access the relevant records in the distributed ledger and simply extract the required information. In fact, because technology companies often rely on seemingly plausible correlations rather than actual and complex human behavior, they may even automatically punish financial institutions that deserve some help or require slight correction. Therefore, they must constantly learn through trial and error to approach regulatory decisions infinitely.

In a multi-objective relationship architecture, there is controversy surrounding the potential consequences of actions and the criteria for collective acceptability of decisions. When pragmatic third-party technology enterprise behavior is an integral part of decision-making, regulatory authorities may be tempted to cross boundaries and obtain benefits beyond their ability to fulfill regulatory tasks in order to achieve regulatory objectives required information. Just as the typical scenario of "panoramic banking" is no longer zero sum competition, but exponential growth of value achieved through extensive cooperation. So, how to adhere to the win-win concept with third-party technology enterprises, complement each other's advantages, and face the uncertainty brought about by innovation? How to encourage all parties involved in regulatory technology cooperation to dare to try and make mistakes, to give certain preferential treatment and protection in terms of performance, to reduce innovation concerns, and to leverage synergies? At the same time, even if the social goals of regulatory authorities are used as indicators, the diversity of policy objectives, difficulties in choosing between policy objectives, obstacles in policy implementation, poor connections in the innovation process, and conflicting starting points and goals among multiple departments can all easily lead to the failure of multi-party cooperation. How to adopt a higher-order inclusive regulatory or governance approach, and how this multi-party participatory decision-making model should be centered around a double-layer fault tolerance mechanism, are all urgent practical questions that need to be answered.

4. Concept transformation and practical shift: with a dual layer fault tolerance mechanism as the core

With the development of financial system reform, experimentation and errors, cost overruns, and abandonment of failed projects have become part of the innovation process, and specific procedural reviews should not be completely disregarded in terms of cost-benefit ratios. Due to regulatory authorities having the power to decide what innovation is, every project must undergo review to ensure proper regulatory technology design to avoid failures in fairness, legality, and credibility. In addition, strict regulatory actions must be distinguished from non innovation and inaction. Acting instead of acting is itself a major failure, and regulatory authorities should still retain the power to revoke authorization and open up regulatory technology, in order to control the depth and frequency of cooperation with each private sector. In other words, the balance fulcrum of the "fault tolerance mechanism" lies in how to implement proactive actions on the basis of defense.

4.1 Concept shift: towards fault-tolerant regulation 

The concept of fault-tolerant regulation comes from a new connotation of inclusive development. Inclusion is an important component of China's excellent traditional culture. Li Dongyang of the Ming Dynasty once said in his "Ode to the Emperor's Elegy": "All plants and emotions are nurtured, and heaven and earth are all inclusive." Su Shi of the Song Dynasty once said in his "Book of Emperor Shangshen Zong": "If Your Majesty is tolerant in multiple ways, then only then can he take second place." Western scholars have directly applied it to business practice, and Otlunde C. A. Johnson has written an article describing the development of "inclusive regulation." Regulators must continuously adapt to the rapidly developing new environment, deepen their understanding of the regulated objects, and achieve civil rights goals such as inclusiveness and equality. In other words, regulatory authorities and policy makers should actively participate in the design of regulatory technology at the beginning of innovation, working together with regulated parties and third-party technology companies to achieve the goal of truly reflecting the intentions of regulatory authorities. The author believes that in order to adapt to the regulatory environment of the digital economy era and solve the regulatory difficulties brought about by financial technology and digital finance, the participation of regulatory authorities does not represent excessive regulation. Greater flexibility should be reserved for the future actions of regulatory technology. After comparing with completely unregulated and sporty regulation, regulators will inevitably change their regulatory thinking and upgrade to a fault-tolerant regulatory model.

4.1.1 Fault tolerant regulation is the regulation that achieves a balance of interests

Fault tolerant regulation is not unregulated, but an attempt to introduce third parties, break existing interest groups, and find new balance points in regulation. For the financial industry, non-regulation is the most easily overlooked major failure. Faced with the constantly changing new characteristics of financial technology, regulators have the power to supervise financial institutions and protect financial consumers. They must abandon conservative traditional regulatory thinking and cannot let new things go unchecked and do nothing because they cannot grasp the trends of new things. The fault-tolerant mechanism encourages regulatory authorities to actively innovate, continuously learn and accumulate experience from failures, and prevent failure to regulate due to fear of making mistakes and accountability. Therefore, fault-tolerant regulation must grasp the following three concepts. One is that fault-tolerant regulation cannot be limited by the complexity of legislation and law enforcement, resulting in a "vacuum zone". Financial legislation itself involves a complex game process, resulting in legislation being unpredictable and historically accidental. In the process of law enforcement, regulators cannot relax or even fail to regulate simply because they do not grasp the right direction but should seek reasonable space limits for fault tolerance. Secondly, fault-tolerant supervision should introduce third-party technology enterprises to implement cooperative supervision at the execution level. Regulatory authorities are at an information disadvantage compared to third-party technology platforms and cannot turn a blind eye to the chaos of subsidies during the start-up period, nor can they be helpless in dealing with the concentration of operators during the adjustment period. When regulators face the highly complex and uncertain new financial formats brought about by the digital economy, the financial risk special rectification design of the " Unidirectional division of labor collaboration" mechanism can no longer adapt to the development of the market. Due to the virtual, non direct contact, and technological properties of digital financial business, regulatory authorities need to verify the identity of the regulated, verify the authenticity of electronic evidence There are practical difficulties in determining transaction behavior and breach of contract liability. Fault tolerant regulation means introducing third-party regulatory technology companies, breaking the limitations of the original scope of separate regulatory functions, not only achieving information exchange between regulatory authorities, but also filling the information gap between regulatory technology and compliance technology. Thirdly, fault-tolerant regulation promotes cooperation between regulatory authorities and social forces, mobilizes third parties, industry associations, the public and other entities, integrates diverse forces, and forms a new inclusive and prudent regulatory model. Increasing reliance on technology will bring new operational risks to financial regulatory authorities, especially in terms of network security and data security. Failure to take strong confidentiality protection measures will damage regulatory credibility and even expose regulatory authorities to litigation risks. Therefore, it is necessary to use the fault-tolerant mechanism of regulatory technology to prevent regulatory authorities from easily abandoning innovation when facing the risks. At the same time, regulatory authorities must actively promote institutional innovation, eliminate their own departmental interests, provide multi-dimensional public services for new financial formats in regulatory strategies, and enhance credibility and execution. In short, fault-tolerant regulation aims to encourage, tolerate, and promote innovation, leaving enough development space for regulating new technological formats, channels, and products, and cannot be simply banned or completely unregulated.

4.1.2 Fault tolerant regulation is a normal regulation that stabilizes expectations

Fault tolerant regulation is not a sporty regulation, but a regular regulation that can be effectively implemented. Sports based regulation is also prone to undeniable consequences of regulatory failure. It naturally has short-term, mobilization, and mandatory characteristics, and has become a regulatory tool repeatedly used by financial regulatory authorities at all levels who pursue a "short, smooth, and fast" effect. It is not a unique phenomenon in China, but widely exists in financial legislation and enforcement both domestically and internationally. Famous scholars assert that the main financial laws are often enacted during the financial or economic crisis, because the temporary anger of voters forced legislators to take legislative action in a hurry, resulting in the formation of many "foam legislation". The most typical representative, Roberta Romano, pointed out that populism and the reaction to the excessive panic of the financial crisis became the main reason for the financial legislation foam. After unpredictable major events trigger financial legislation, policy entrepreneurs who lobby for new legislation in the corridors of power represent and respond to changes in the endogenous dynamics of social, political, and economic forces, but rarely grasp substantive local legislation to help the government implement new financial legislation. In other words, the history of financial development is already a product of transactions between financiers, populist activists, or other independent groups. But these opportunistic "policy entrepreneurs" consider their "outdated" policy recommendations as "solutions" to current emergency situations, and rarely consider the practicality of these suggestions. As people's memories gradually fade away, legislative and regulatory authorities shift their attention to other areas, and regulators will use this negligence to influence regulatory agencies, weakening their efforts in rule making and enforcement. The hasty and passive nature of "sports style legislation and regulation", as well as the rebound brought about by the rectification results, and the possible disruption to the principle of fairness in the rule of law, have led to criticism of its regulatory performance. Therefore, in the era of digital economy, the regulatory approach of shifting from left to right should be abandoned, and a normal and fair regulation with fault-tolerant mechanisms should be called for to smooth out the economic cycle. Normalized implementation of fault-tolerant supervision can provide regulated institutions with relatively stable regulatory expectations, aiming to guide their own standardized development, encourage regulatory technology cooperation parties to dare to try and make mistakes, and provide them with appropriate space and flexibility.

4.1.3 Fault tolerant regulation is innovative regulation that follows legal procedures

Fault tolerance regulation is not a regulation that departs from procedural justice, but should provide relative fault tolerance space according to legal procedures to promote the realization of "sustainable development of the overall social interests" in regulation. As Li Keqiang said, if we continue to use the old methods of regulation, companies like WeChat may no longer exist. Promoting fault-tolerant regulatory technology not only avoids stifling market innovation, but also effectively connects regulatory agencies with regulated institutions through regulatory and compliance technologies, allowing regulatory technology to emerge and develop to the maximum extent possible, and reserving flexible and effective regulatory fault-tolerant space. The internal model method for capital adequacy ratio regulation should be regarded as a new governance technique, which involves incorporating the internal capital model of financial institutions into the capital adequacy ratio system, providing real-time supervision during the reporting of model data, and appropriately compensating for the information asymmetry defects caused by the complexity of contemporary financial institutions. Although regulatory technology governance tools have multiple advantages when applied in highly complex and dynamic regulatory contexts, their internal model methods are susceptible to technical interoperability and literal interpretation traps between different financial regulators and regulated entities, which may compromise the legitimacy and effectiveness of democracy. Therefore, in the context of social public interests and the overall national security concept, principle-based supervision should be used instead of rule based supervision as much as possible, in order to provide third-party technology platforms with the space and time to independently apply for regulatory technology and compliant technology projects, as well as to explore regulatory technology and compliant technology market standards and rules. Regulatory authorities should provide services and testing environments according to due process requirements, and may issue universal guidance standards, leave enough space for market entities to choose project testing methods according to their own needs, allowing regulators to independently determine the scope, types, methods, and other content they need to try. For the rigid idea that regulating technological innovation may cause harm to the core public interest, it is necessary to construct a "fault-tolerant legal relationship", which is a social relationship with rights and obligations formed in the process of regulating technological innovation activities. Based on this, a fault-tolerant and error correction legal mechanism must be constructed to protect both the public interest and the rights and interests of decision-makers in entrepreneurship and innovation. Hobhouse once pointed out that "various social and political systems are not goals in themselves. They are organs of social life, and whether they are good or bad depends on the spirit they contain." In the pursuit of regulatory technological innovation, regulatory agencies often enrich the innovation of systems and intentionally or unintentionally reduce the requirements for legitimacy standards, The fault-tolerant mechanism is likely to become a breeding ground for violations to evade punitive regulatory enforcement during implementation. Therefore, consumer rights and data rights should not be swallowed up by regulatory technological innovation, nor should true innovation be hindered by choking on food. Instead, fault-tolerant innovation should be carried out while maintaining smooth communication between regulatory authorities and third-party technology enterprises, adhering to the legal basis of financial regulation and compliance system design.

4.2 Practice shift: rational application and optimization of regulatory technology

The regulatory technology fault-tolerant mechanism is an important lever for the "innovation driven development strategy" and a specific proposition in financial inclusion. The purpose of constructing the upper level fault-tolerant mechanism is to examine the function and positioning of regulatory power in the development of the digital economy market. Financial regulators are the basis for the legitimacy of regulatory theory in the trilateral interaction between regulatory technology, leaving ample room for trial and error in both regulatory technology innovation and financial technology innovation, We need a fault-tolerant space that allows for failure to relax restrictions on regulatory innovation, while also improving the efficiency of cooperation between the government's intervention and market regulation hands, finding a "third path" that goes beyond the binary opposition between the government and the market, and encouraging the integration and innovation of regulatory and market entities.

4.2.1 Main body determination and system compatibility of fault tolerance mechanism

A dedicated research institution should be established to promote fault-tolerant supervision in collaboration with the three main objectives. The subject of the fault-tolerant mechanism refers to the relevant subjects that initiate the fault-tolerant error correction program. In the dual fault tolerance mechanism, the Central Financial Commission (hereinafter referred to as the Central Financial Commission) and the Central Financial Working Committee (hereinafter referred to as the Financial Working Committee) are the upper level fault tolerance entities. For major errors, it is necessary for the disciplinary inspection and supervision department or the personnel department to review and determine whether there are errors and whether they are tolerable based on the prescribed detailed procedures independently or jointly. If there is no error or tolerable error, the result can be handed over to the Central Financial Commission and the Financial Working Committee, and they will declare the handling of the error as no error or tolerable error. According to the March 2023 Plan for the Reform of Party and State Institutions, the Central Financial Commission is the decision-making and coordination body of the Party Central Committee. It will strengthen the Party Central Committee's centralized and unified leadership over financial work, responsible for top-level design, overall coordination, overall promotion, supervision and implementation of financial stability and development, and studying and reviewing major policies and issues in the financial field; the Office of the Central Financial Commission, as the administrative agency of the Central Financial Commission, is included in the sequence of central party institutions. Its functional positioning has the legality and suitability of identifying fault tolerance situations and initiating fault tolerance exemption work. Regulatory technology is the process of comprehensive technological transformation of financial regulatory functions. The object of fault tolerance and exemption in the upper-level mechanism is the vast number of financial regulators, and it is necessary to encourage regulators to take on responsibilities and innovate in fault tolerance; The improvement of the upper level fault tolerance mechanism will promote the smooth development of financial technology fault tolerance mechanisms. The two interact and blend seamlessly, and the country completes this major mission by creating or authorizing specialized regulatory agencies.

In addition, in the latest financial system reform in China, the function of systematic financial prevention belongs to the People's Bank of China, while the State Administration for Financial Supervision and Administration is responsible for overall protection of the rights and interests of financial consumers; in this way, the People's Bank of China and the State Administration for Financial Supervision and Administration of China have become truly the leading state organs of the lower level regulatory technology fault-tolerant mechanism. Regulatory authorities, as key entities in the dual fault tolerance mechanism, are crucial for regulatory technology in pursuing and achieving financial stability and consumer protection goals. However, no matter how clear the regulatory objectives and governance system are, regulatory rules cannot cover all regulatory technology scenarios and have uncertainty. The subjects in the fault-tolerant mechanism should grasp the corresponding residual power space. On the one hand, regulatory authorities must distinguish which technical faults in auxiliary systems belong to "normal errors" and which appear to be "minor faults" but may cause a series of unexpected cascading effects, and even significant financial risks. On the other hand, the disruptive and innovative nature of financial technology and regulatory technology has led to a sharp expansion of this residual power space, which is difficult to handle based on the inherent experience of regulators. Usually, regulatory authorities constantly weigh the scale and allocation of remaining power space in financial regulation. The "agreement" between regulatory authorities and regulated entities or third-party technology companies under seemingly established constraints is the result of financial regulatory authorities occupying a dominant position and continuously measuring in the process of power allocation. The difference lies in the shift of regulatory technology from preadmission regulation to mid to post event regulation, emphasizing the establishment of a multi-party regulatory technology coordination and cooperation mechanism, setting fault-tolerant unified market standards, and improving consumer protection mechanisms for fintech.

The gradual nature of China's financial reform determines that the dual fault tolerance space of regulatory technology will continue to exist, forming a dual fault tolerance mechanism of regulatory technology as shown in Figure 1. The error supervision and remedy in the regulatory technology fault tolerance and correction mechanism may be self-recognized and corrected by financial regulatory authorities, which means there is a certain margin of fault tolerance. If regulatory objectives, regulatory capabilities, the behavior of regulated parties and third-party technology enterprises change, the above "consensus" state will inevitably be broken, and regulatory authorities must readjust the configuration of the fault tolerance mechanism. In this way, only beyond the design space of fault tolerance mechanisms, other departments such as decision-making authorities, disciplinary inspection and supervision departments, or judicial authorities need to define and supervise.

Figure 1: Dual layer fault-tolerant mechanism of regulatory technology

4.2.2 Innovative development of fault-tolerant mechanisms and dual layer content of risk prevention

The fault-tolerant mechanism has shifted from simply developing regional experimental fields to promoting the dual development of regulatory technology innovation and risk prevention. The Party regulations at the central level stipulate that "errors in promoting reform due to lack of experience and prior experimentation, errors in exploratory experiments without clear limitations, and unintentional negligence in promoting development" belong to situations where accountability is not held or exempted, leaving room for fault tolerance in the upper level of regulatory technology. One effective way to reduce legislative errors in the field of financial regulation in the lower level of regulatory technology is to allow for regulatory experimentation. Among them, "what is wrong to tolerate" is the reason for the initiation of fault tolerance and exemption, and in essence, it is the problem of defining the "fault tolerance situation". On the one hand, the "errors" that are tolerated in the dual layer fault tolerance mechanism are still in the early stages of development. In order to promote the digital transformation of the financial industry and its regulation, it is necessary to scientifically set up "observation periods" and fault-tolerant conditions for emerging financial products or formats, implement an objective "no punishment for first violation" system in accordance with the law, and explore the establishment of a financial technology regulatory model that adapts to the digital economy era. If the only consequence of a failed regulatory technology innovation process is a waste of time and resources, then such failures need to be tolerated the most, as regulatory authorities may learn important lessons from failed innovations. On the other hand, to prevent one-sided tolerance towards financial innovation, it is necessary to deeply analyze the risks that financial innovation may bring. The regulatory requirements for traditional financial activities and financial innovation activities should be treated equally, and gross negligence, intentional violations of the law or enforcement bottom line should not be tolerated. Unrestricted incentives will inevitably lead to fault-tolerant legal relationship entities pursuing their own interests and damaging public interests. It is particularly noteworthy that cyber-attacks often target the infrastructure that financial regulatory authorities rely on, which may paralyze their ability to fulfill their supervisory duties. Therefore, regulatory authorities need to repeatedly experiment and innovate through regulatory technology, adopting methods such as access control, user authentication, data encryption, and strong firewalls to resist internal and external threats.

Technological innovation is advancing rapidly, traditional finance is constantly changing, and society is also experiencing a phenomenon of pan gold melting. New forms of objects are constantly emerging, and new businesses, models, and formats are emerging, constantly creating new social relationships. Objectively, it requires financial laws to constantly innovate and trial and error, expand the scope of adjustment, avoid regulatory blind spots, and gradually provide institutional guarantees for financial technology and regulatory technology innovation through confirmation, adjustment, revision, and other means. In a sense, China's financial legal system has developed and improved through a trial and error, error tolerance, and correction process, starting from point to surface and then covering all aspects. Especially, the development of financial technology in China provides an opportunity for trial and error in financial regulation. Regulatory authorities can proactively and reasonably relax regulatory regulations, reduce rule barriers, and determine the scope of regulation based on the final test results. In fact, regional pilot fields and "experimental supervision" effectively isolate financial transmission risks, that is, intervene in a timely manner when one or several components fail, and place potential errors within a controllable range to avoid triggering more component linkage effects. Therefore, it is crucial to consider the failure of inaction as a failure and categorize it within the scope of non tolerance, as inaction in regulating technological innovation means jeopardizing the fundamental goals of financial stability and consumer protection.

From the perspective of the relationship between the upper and lower fault tolerance mechanisms, if regulatory technology fails to achieve fair distribution of regulatory costs and benefits, then the fault tolerance of financial technology will inevitably be unsustainable. In the context of an exceptionally large, rapidly transmitting, and highly interconnected social information ecosystem, financial markets are facing potential cascading failures on a global scale, and there is an urgent need for regulatory technology to conduct fault-tolerant experiments.

Firstly, regulatory technology should evolve synchronously with the fault-tolerant mechanism of financial technology. Since 2019, to implement the "FinTech Development Plan (2019-2021)", the People's Bank of China, together with the National Development and Reform Commission, the Ministry of Science and Technology and other departments, has organized pilot projects for the application of FinTech in provinces and cities such as Beijing, Shanghai, and Zhejiang. These pilot projects have been approved under the supervision of regulatory authorities such as the People's Bank of China, and traditional methods have not been simply applied. However, there is still significant room for improvement in the development of emergency and exit mechanisms in advance, which can allow regulators to intervene early in the judgment of whether an innovation is truly innovative, improve the innovation triggering and promotion process of regulatory authorities, timely adopt measures such as risk provision funds and insurance plans, establish a multi-level and three-dimensional risk control system, and create a flexible and efficient trial and error tolerance mechanism.

Secondly, in the process of integrating regulatory technology with compliance technology, the focus of fault tolerance should be clarified. At present, the management of some financial regulatory "trial" and "trial" legislation in China is still relatively extensive, and some documents do not even clarify the core points of "trial and error", lacking more consideration of the rights and interests of relevant parties, resulting in difficulties in protecting rights such as the right to information and privacy. When dealing with problems such as algorithmic black boxes or algorithmic discrimination, most regulated objects or third-party technology companies find it difficult to fully judge the fairness of algorithms, and regulatory authorities should take on regulatory responsibilities at this moment. On the one hand, if the decision-making power of automatic circuit breakers such as freezing, transferring, and cutting off fund supply is granted to machine learning algorithms, it may affect the vital interests of market entities. Therefore, regulatory authorities must maintain a cautious attitude, repeatedly experiment, and retain the power of regulatory review; On the other hand, while improving the flexibility of regulatory technology, maintaining continuous regulatory analysis and risk assessment, establishing a positive interaction and information exchange between regulators and regulatory technology enterprises, implementing flexible, adaptive, and differentiated supervision at different levels, leaving enough room for error tolerance, and reducing the game cost of regulatory execution.

Thirdly, regulatory measures for technological innovation products should also include arrangements for exemption mechanisms. A buffer zone can be established before administrative penalties, providing necessary toolboxes, mainly for compliance education and rectification, and providing clear legal basis for illegal accountability, including clear exemption conditions for regulatory accountability and compliance accountability. In this way, regulators adopt a tolerant strategy around regulatory technological innovation and its inevitable setbacks, inspiring them to abandon short-term and regulatory arbitrage behavior.

Fourthly, the cooperation between regulatory authorities and third-party technology platforms has dual fault tolerance. On the one hand, regulatory authorities have sufficient rationality and legitimacy in both the "new" dimensions of rights and the "empowerment" of interest claims, and this is an emerging right centered on the "public subject". Therefore, based on the principle of equality and guided by top-level design, law enforcement officers who make the same mistakes should be given equal protection. On the other hand, collaborating to develop regulatory technology tools or products requires better incentives for fintech companies to improve their service models, avoid business risks, and encourage regulatory authorities to introduce regulatory rules more accurately. According to the "four no's" judgment criteria applicable in fault-tolerant policies, namely "whether regulatory technology violates policies and regulations, whether it follows decision-making procedures, whether it has actual consequences, and whether the three parties have colluded to make profits." Determine the different conditions that can, should, and prohibit fault tolerance, and consider the scale of leniency, mitigation, and exemption of fault tolerance, in order to alleviate the regulatory competition, rule obsession, short-term ism, regulatory capture, and other issues caused by the lagging regulatory technology mentioned earlier. Regulatory authorities should learn from their mistakes, and deployed technological innovations should also be adjusted according to the emergence of new technologies or the constantly changing environment.

4.2.3 Program priority and accountability precision of fault tolerance mechanism

When there are errors in the substantive regulatory technology access review, even if the behavior of the actor has been "wrong" and meets the legal conditions, the regulatory authorities and third-party platforms can raise objections. If the error cannot be attributed to the actor, the actor does not need to bear corresponding responsibilities or should be relieved or reduced in liability. The regulatory authorities can still make targeted repairs; When there are errors or defects in the procedural admission review, there is no opportunity or space for repair. From this perspective, in the context of regulating technological innovation, public interest is also a legitimate procedural issue.

The first point is to use efficiency fault tolerance as an independent prerequisite for initiating accountability. Firstly, a selection and screening process should be established. Given the multidimensional nature of regulatory rules designed for financial business, natural language processing inevitably reflects the views of data scientists on the importance (or non-importance) of certain elements in the text. The steps taken may include filtering common or uncommon words, deleting numbers, symbols or proprietary names, etc.. If data scientists exclude elements of significant importance in financial prudential regulation in program design, natural language processing analysis will naturally make errors due to this defect. Even if good projects are selected to enter the "exemption zone" after human supervision or review and evaluation of program errors, it still cannot guarantee the improvement of regulatory efficiency; For those who do not comply with financial exemptions, application and objection channels should be provided. If the application or objection is successful, they can enter the "regulatory sandbox" or "experimental exemption zone" for testing, maximizing the role of fault tolerance mechanisms.

The second is to strengthen the construction of a programmatic regulatory system with fault-tolerant mechanisms. Establishing differentiated and personalized regulatory tools and processes for different types of new risks may still outperform traditional regulatory measures, even with imperfect regulatory technological innovation. By establishing emergency exit and trial and error tolerance mechanisms, we will improve risk compensation measures. Relying solely on manual and scattered supervision methods such as window guidance and off-site inspections is far from keeping up with the leapfrog development of the times. It is urgent to strengthen the construction of regulatory technology capabilities with fault-tolerant mechanisms. Regulatory authorities should adopt technological means to achieve standardization, digitization, and routinization of regulatory rules, strengthen the depth and breadth of regulatory technology penetration, establish a procedural regulatory technology rule system and application norms with fault tolerance mechanisms for security risks caused by innovative applications of financial technology, conduct trial and error innovation within a controllable range of risks, and improve the information approval system and dispute resolution mechanism in trial and error tolerance. And the specific scope of application, supporting mechanisms, division criteria and standards, and exemption procedures.

The third is to strengthen the pre filing, post interpretability, and precise accountability system with a fault-tolerant mechanism. In the highly technological and open responsive legal perspective, the status of legal responsibility will be more prominent. If the responsibility mechanism is too rigid, it will damage the openness of the legal system, promote formalism and retreat; similarly, granting broad discretion is a prerequisite for the preservation of openness in the law. If the responsibility mechanism is not strict enough, there is a risk of openness and flexibility deviating into non-binding. On the one hand, for the upper-level fault-tolerant mechanism of regulatory innovation, a full process "traceability plan" should be implemented for the innovative decision-making and trial and error behavior of regulators. Timely feedback on the progress of innovative decisions that have already been filed. Once deviations in the implementation of decisions are found, corresponding measures should be taken immediately to prevent serious consequences. The Central Financial Commission and the Financial Industry Commission shall give a determination opinion, and if necessary, a hearing procedure may be introduced. They may also consult with other departments such as national data, online information technology, and industry and information technology to make a decision. If the decision is difficult or may cause significant harm and impact, it shall be reported to the State Council if necessary; the disciplinary inspection and supervision organs should promptly clarify any untrue issues, and hold accountable the regulators for their innocence. For truthful issues, they should be held accountable in accordance with relevant regulations and urge the corrected parties to actively correct and stop losses. On the other hand, in the lower-level fault tolerance mechanism of regulatory technology, emphasis is placed on the post interpretability of technology. When financial regulation is achieved through technological means, behavioral choices and value judgments are naturally embedded in the entire technological design, in a sense, consolidating regulatory compromises reached at a certain moment. Technological forces may be more automatic and self-executing than traditional regulatory strategies, easily amplifying deviations in behavioral choices and value judgments. The Evaluation Specification for Financial Applications of Artificial Intelligence Algorithms puts forward clear interpretability requirements for the application of algorithms in the financial field, especially the use of post interpretability in black box models. Afterwards interpretability refers to the use of interpretable methods such as SHAP for black box models, explaining the decision-making process and sensitivity relationship between input and output of the black box model, and increasing the trust of model users. Interpretability after the fact means that specific "errors" should be allowed or even encouraged in advance to integrate rigid responsibility mechanisms into moderate fault tolerance mechanisms, in order to meet the functional enhancement effect of responding to innovation and encouraging responsibility at that time. In addition, it is necessary to promote the rational and precise development of accountability. The background of strengthening accountability reform has given rise to a negative perception of "doing more, making fewer, and making fewer mistakes", resulting in some financial regulatory officials being subjectively hesitant or unwilling to act, and not taking responsibility or action. Precision accountability is an important measure to encourage regulatory authorities to shift from "avoiding responsibility and not doing anything" to "actively taking action", and is an inevitable remedy for tolerance empowered fault tolerance mechanisms. For those who comply with laws, regulations and policies, are inevitable for innovative development, do not seek personal gain, have no subjective intention, and have taken necessary measures to prevent or correct errors before or after their occurrence, they can be recognized through certain procedures and exempted from relevant responsibilities.

Not allowing exploratory errors in reform and innovation is undoubtedly putting a "tight curse" on reform and innovation practitioners, hindering the progress of comprehensive deepening reform. Fault tolerance and error correction is a systematic engineering project. In the financial technology and regulatory technology fields with typical cross industry, cross-border, and cross domain characteristics, fault tolerance mechanisms neither affirm errors nor avoid responsibility. They should not lead to the loss of space for criticism and improvement of the law, but also prevent the occurrence of "evil governance" under the name of "legality". The emergence of new contradictions and problems in regulatory technology under the new normal of the digital economy calls for the construction of a fault-tolerant mechanism to ensure the smooth progress of comprehensive deepening reforms in the financial regulatory field.

5. Conclusion

Good regulatory scientific and technological innovation should not only repeat the history of Internet finance, but should deeply reflect on the historical roots and legal basis of inclusive regulation, covering three dimensions: the goal of Chinese path to modernization financial regulation, the compliance motivation of regulators, and the innovation motivation of third-party technology enterprises. The two-tier fault-tolerant mechanism should be regarded as the core of the current new financial regulation mechanism composed of "key technology+institutional guarantee". One is technical screening, which is a regulatory technology that can extract and distinguish systematic financial risk clues from chaotic information. Regulatory technology can assist regulatory authorities in summarizing and analyzing data from regulated institutions and can also combine the risk trends reflected with external big data. By using artificial intelligence tools such as predictive and emergent big models, potential sources and signals of systemic risks can be identified before widespread risk damage occurs. The second is legal incentives. Legislators and regulators must have sufficient awareness of fault tolerance and provide a relaxed environment for regulatory and financial innovation to experiment with new ideas through the improvement of rules and regulations. Regulatory rules should include both prohibitive and mandatory rules, as well as authorization and facilitation rules. Thirdly, the concept is inclusive, viewing regulation as a flexible and differentiated public service provided by the government with fault tolerance. Promote knowledge generation channels for addressing uncertainty issues as much as possible, always pay attention to and study corresponding potential outcomes, and update regulatory technology through feedback, making regulatory technology highly compatible with fault tolerance mechanisms. Fourthly, it is feasible in practice to establish a "option centered" regulatory technology fault-tolerant mechanism, and to screen out excellent fintech enterprises and regulatory technology projects as much as possible. Both supporters of sandbox regulatory technology and opponents of a universal standard rule system should embrace the fault-tolerant mechanism of regulatory technology in order to maximize the screening of excellent enterprises and projects. In essence, acknowledging the fallibility of practice is acknowledging the fallibility, exploratory, open, and unfinished nature of theories derived from practice. Inclusive regulation, with a focus on fault-tolerant mechanisms, focuses on actively shaping market order and regulation. Through multi-party cooperation, data sharing and integration are achieved, data-driven regulation and algorithmic regulation are established, ultimately achieving the legitimacy, effectiveness, and authority of prudent regulatory technology. Based on maintaining the bottom line of preventing systemic financial risks, financial technology innovation is promoted.

The fault-tolerant mechanism is an institutionalized manifestation of Marxism's advocacy for practical spirit. The specific manifestation of fault-tolerant thinking may undergo necessary changes from computers to aviation, industrial control, medical, and regulatory technology fields, but the internal mechanism will still remain relatively stable. The dual fault tolerance mechanism is not omnipotent, but it is a necessary condition for the long-term, sustainable, and innovative development of regulatory technology. It is reasonable to adopt a tolerant attitude towards the mistakes made by pioneers in exploration, but building a fault-tolerant mechanism still has a long way to go. How to truly establish the internal mechanism, operating system, and procedural norms of technology driven regulatory fault-tolerant mechanisms, how to determine whether the power of innovation is mismatched, how to share the cost of trial-and-error correction, and who determines the exemption and accountability of regulatory technology innovators and errors within the current legal framework still need further practical development and corresponding in-depth research.

The original text was published in the "Practical Research" column of the first issue of “Politics and Law” in 2024. Thanks for the authorized reprint of "Politics and Law" on the official account of WeChat.