A Regulator Born from Conflict?

The idea of creating an AI self-regulatory body modeled after FINRA has gained momentum. Google DeepMind CEO Demis Hassabis champions the proposal, which would establish a voluntary system where leading AI labs fund the new organization. This body would develop assessment protocols for frontier AI models, conduct safety and security testing, and encourage companies to adopt governance standards.

The involvement of prominent figures like Microsoft’s Satya Nadella, Block’s Jack Dorsey, and Elon Musk lends credence to the proposal. Even David Sacks, former Trump administration AI czar, has expressed support. However, concerns arise that this self-regulatory body would suffer from the same conflicts-of-interest issues plaguing FINRA.

The Problem with Self-Regulation

FINRA’s experience serves as a cautionary tale. Despite its best intentions, the regulator often prioritizes the interests of brokerage firms over individual investors. Whistleblowers have accused FINRA of failing to address problematic brokers adequately and handing out fines that are too low. This pattern raises concerns about an AI self-regulatory body’s ability to police companies.

Nader Henein, a VP analyst at Gartner, argues that self-regulation is not viable due to conflicts of interest and the limited capacity of tech vendors to regulate themselves. The reliance on voluntary submissions from AI companies also questions its effectiveness. Hassabis suggests that once evaluations are shown effective, the system could become mandatory for any model distributed in the US.

However, this assumes cooperation from companies whose interests may not align with those of the public. History shows self-regulatory bodies often turn a blind eye to problematic practices in favor of maintaining relationships with powerful corporations.

The Limits of Voluntarism

The proposed system’s voluntary nature raises questions about its effectiveness. Hassabis suggests that once evaluations are shown effective, the system could become mandatory for any model distributed in the US. However, this assumes cooperation from companies whose interests may not align with those of the public.

Critics argue that FINRA has often prioritized relationships with major financial firms over addressing problematic practices. This pattern raises concerns about an AI self-regulatory body’s ability to regulate effectively.

The Risk of Normalization

The enthusiasm surrounding this proposal masks a concerning reality: the normalization of conflict-of-interests in regulatory bodies. By creating an AI self-regulator modeled after FINRA, we risk perpetuating the same problems that have plagued traditional finance.

As we move forward, it’s essential to critically evaluate the implications of such a system. Will the involvement of prominent figures and voluntary submissions ensure accountability? Or will this new regulator become just another example of how powerful corporations can game the system?

The answers won’t come easily, but one thing is certain: we cannot afford to repeat the mistakes of the past.