Kevin Frazier (The University of Texas School of Law) has posted Move Fast and Delegate Things: Can a FINRA for AI Survive Constitutional Scrutiny? on SSRN. Here is the abstract:
Highly-capable AI models present risks that warrant heightened scrutiny. These risks include but are not limited to cybersecurity threats, bioweapon development, mass surveillance, and loss of control scenarios in which models escape their test environments or otherwise bypass imposed safeguards. Recent testing by the UK AI Safety Institute documenting how frontier models may pursue long-term goals via unsanctioned behavior bolsters the case that such risks are prevalent and, under present safeguards, inadequately mitigated.
Existing government systems are not suited to detecting those risks nor to evaluating the adequacy of mitigations created by the labs. This lack of capacity is the result of several commonly accepted observations: relative to the labs, the government has limited AI expertise; given the current and forecasted pace of AI progress, regulatory processes move too slowly to respond to new developments in AI or may be applied in an ad hoc, arbitrary fashion; in expectation of those AI advances, there’s a corollary need for investments in the science of AI safety, including how best to evaluate models; the scale of those investments as well as the research ecosystems necessary to advance the science of AI are not well-suited to current government institutions; and, in alignment with a general desire for US models to be the default models around the globe, an independent system of AI evaluation may carry greater odds of acceptance by other nations than a process solely operated by the US government.
These observations amount to a serious argument for delegating substantial regulatory authority to private experts. Yet the argument is incomplete. Audited self-regulation has a documented history, and that history includes recurring failure modes and constitutional barriers that its advocates may not fully acknowledge.
Recommended!
To receive new posts from Legal Theory Blog by email, get a free subscription to Legal Theory Stack.
Lawrence Solum