US Pursues FINRA-Like Approach to AI Regulation

Is the "Wild West" of AI Ending? A FINRA-Like Independent Oversight Initiative Emerges from the US!

The artificial intelligence revolution is progressing at an unprecedented pace, bringing with it an escalating tide of uncontrolled power and potential risks. What were once subjects of futuristic scenarios—self-learning systems—have now infiltrated every aspect of life, from financial markets to cybersecurity, and even strategic decision-making. Yet, beneath this immense potential, these autonomous systems and complex Agentic Workflows, operating without sufficient oversight and transparency, are creating a veritable "Wild West."

Amidst this chaotic landscape, a historic move is emerging from the US: the creation of an independent oversight body for AI models, mirroring the FINRA model in the financial sector. This development not only has the potential to shape the future of global AI governance but could also signal the end of the "self-regulation" era. Developed by key figures like Treasury Secretary Scott Bessent and White House Chief of Staff Susie Wiles, this plan is a significant proposal anticipated to be implemented under a potential Trump administration. Strongly supported by Google DeepMind CEO Demis Hassabis, this vision could herald a new era in AI oversight. Let's delve into the backstory of this groundbreaking development.

The FINRA Model: A Guiding Star for AI?

So, what is the FINRA model, and why is it so appealing for artificial intelligence? The Financial Industry Regulatory Authority (FINRA) is an independent self-regulatory organization (SRO) in the US that oversees brokerage firms. It is funded by the industry but operates under the oversight and authorization of a powerful government authority like the Securities and Exchange Commission (SEC). Its purpose is to protect market integrity and investors. This dual structure—the industry's own expertise and funding combined with the government's ultimate oversight and enforcement power—forms the bedrock of FINRA's credibility.

"Even Silicon Valley leaders are vocal about the 'disorderly' nature of US AI initiatives, loudly articulating a need for regulation. This is just the tip of the iceberg."

The primary reason for adapting such a structure to the AI world is to ensure that advanced 'frontier' AI models are independently evaluated for critical areas such as safety, cybersecurity, biological risks, manipulative capabilities, and ethical biases before they are released to the market. Given the potential for AI systems to exhibit 'emergent behavior' and their 'black box' nature, such independent and competent evaluation is vital. Industry experts emphasize the meticulous preparatory work required to ensure that complex autonomous systems and Agentic AI solutions operate within a safe and ethical framework; this is not just about regulatory compliance but also about building public trust.

Even Silicon Valley leaders' discomfort with the 'disorderly' nature of US AI initiatives and their vocal demand for regulation underscore how widespread this search is. The White House's earlier statement that there would be "no FDA for AI" and the previous Trump administration's "laissez-faire" approach to AI essentially made this new model search inevitable. Indeed, in a world where autonomous systems can make their own decisions and execute complex 'agentic workflows,' it became clear that purely voluntary oversight would be insufficient.

Demis Hassabis Takes the Stage: A Call for Global Oversight

One of the strongest proponents of this idea is Demis Hassabis, CEO of Google DeepMind. Hassabis advocates for the establishment of a US-led global AI observer institution. This body, much like FINRA, would comprise independent experts and representatives from the open-source community. Imagine the world's brightest minds coming together to scrutinize the most advanced AI models before their release.

According to Hassabis, this institution would be responsible for subjecting 'frontier' AI models (i.e., the most advanced, powerful, and potentially risky artificial intelligence systems) to in-depth safety tests, bias analyses, and ethical evaluations before they are deployed. Given these models' capabilities, especially their capacity to design 'autonomous systems' and 'Agentic Workflows' that can undertake complex tasks without human intervention, preventing potential errors or malicious uses becomes critical. The institution could also have the authority to slow or halt the release of risky models across the industry.

But how realistic is this vision? Hassabis has been lobbying for this proposal for months, holding discussions with a potential Trump administration and European officials. And according to reports, he has received "very positive" feedback from the White House! Treasury Secretary Scott Bessent is known to be developing the proposal, and White House Chief of Staff Susie Wiles is currently reviewing the plan. Hassabis hopes this organization will be operational by year-end.

The "Grading Your Own Homework" Dilemma: Independence and Trust Issues

As with every great idea, this proposal has its controversial aspects. The possibility that a FINRA-like model, especially as Hassabis initially proposed, would be funded by the very companies it oversees and operate without a superior regulatory "principal" (a high government authority like the SEC) has been criticized with the analogy of "the industry grading its own homework." This is akin to football teams choosing and paying their own referees. How impartial can it truly be?

"There are serious questions about how impartially an organization can oversee the entities that fund it. FINRA's success is underpinned by the existence of a higher authority like the SEC."

As The Economic Times has also pointed out, the fundamental factor behind FINRA's success in the financial sector is its operation under the supervision of a strong and independent public authority like the Securities and Exchange Commission (SEC). The SEC approves FINRA's rules, oversees its activities, and intervenes when necessary. Without this hierarchical structure, it would be difficult for a similar AI institution to avoid "self-regulation" criticisms and ensure genuine credibility. Given the complex internal workings and transparency challenges of AI models, the existence of an independent oversight mechanism and a superior authority is critically important for both trust and accountability.

The US's caution in grappling with legislative and jurisdictional debates on this issue becomes even more apparent when compared to China's plans for more centralized and state-backed recall mechanisms and stringent regulatory frameworks for autonomous software. China views AI technologies as a strategic national priority, while simultaneously implementing strict, state-controlled regulations across a wide spectrum, from algorithmic discrimination to data security. This raises the question of who can advance faster and more effectively in the global race for AI oversight leadership.

Why Now? AI Risks Are on Our Doorstep!

So, why is such a structure needed right now? Because AI risks are no longer theoretical; they're on our doorstep and even through the door. Advanced 'frontier' AI models concretely reveal potential dangers:

• Agentic AI and the Storm in the Financial Sector: Wall Street banks are rapidly increasing their use of 'Agentic AI' (agent-based artificial intelligence) to boost productivity and operational efficiency. Agentic AI systems are capable of planning actions, making decisions, and executing complex tasks autonomously to achieve predefined goals. These systems can operate across a wide range of functions—from optimizing financial transactions and conducting risk analyses to implementing complex trading strategies—without requiring human intervention. Giants like Morgan Stanley, Goldman Sachs, JPMorgan, and Citi are integrating this technology into many areas, from transaction accounting to client screening and even market predictions. However, this situation brings serious concerns regarding providing broad and autonomous access to systems, potential 'emergent behavior' risks, and accountability. In the event of an Agentic AI making an unexpected error or falling victim to a malicious attack, 'runaway agent' scenarios that could cause chain reactions in financial markets and create systemic risks become a real threat.
• Cybersecurity Threats: Advanced 'frontier' AI models like Anthropic's Claude Mythos can not only generate text but also create complex cyberattack scenarios, identify vulnerabilities, or produce manipulative content, thereby escalating cybersecurity threats. If such models fall into the hands of malicious actors, they could amplify the capacity for automating disinformation campaigns or sophisticated cyberattacks. These potential dangers prompted Canadian regulators (OSFI) to act and led US Treasury Secretary Scott Bessent and then-Federal Reserve Chairman Jerome Powell to call an urgent meeting with bank CEOs to warn them about these threats. These events strikingly highlight the potential and tangible dangers of AI models on the 'real world.'
• What About the White House's Current Solutions?: A potential Trump administration plans to launch an information-sharing hub called "Gold Eagle" for AI-driven cybersecurity risks. This initiative offers a platform where companies can voluntarily submit their models for evaluation. However, in an era where 'Agentic Workflows' and 'Autonomous Systems' are rapidly proliferating, questions arise about whether voluntary approaches can provide sufficient protection against potential 'black swan' events or 'zip bomb error' scenarios. A more comprehensive, binding, and independent oversight mechanism appears essential for both fostering innovation and preventing potential catastrophes.

A Critical Juncture for the Future of AI

The US's consideration of establishing a FINRA-like independent body for AI oversight is one of the most concrete and ambitious steps to civilize this 'Wild West' environment. This proposal represents an effort to find a delicate balance between the industry's tendency towards 'self-regulation' on one hand, and the growing need for government oversight in the face of rapidly increasing potential AI risks on the other. This balance is built on ensuring safety without stifling innovation and maximizing AI's societal benefits while minimizing its potential harms.

While strong support from industry leaders like Demis Hassabis gives this initiative momentum, fundamental questions regarding the institution's true independence, the transparency of its funding model, and its oversight by a high-level public authority like the Securities and Exchange Commission (SEC) must be satisfactorily answered. Otherwise, this structure could merely be perceived as "players paying their own referees," losing credibility with both the public and the AI ecosystem, which would ultimately undermine trust in AI in the long run.

Developments in this area in the coming period could be a critical turning point that will shape the future of global artificial intelligence governance and technology, not just in the US. Realizing the boundless potential of artificial intelligence responsibly is the shared duty of not only regulators but also technology developers, researchers, and society as a whole. Let us not forget that whether these technologies become a blessing or a threat to humanity depends on how we manage them today.