White House Announces "Substantial Progress" on AI Executive Order Directives
Commentary by Steven Lofchie
The White House announced that federal agencies completed their ninety day directives set out in President Joseph R. Biden's Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence ("EO").
As previously covered, the AI directives and standards listed in the October 30, 2023 EO were intended to (i) protect citizens against potential technology risks while promoting and advancing competition, (ii) advance equity and civil rights, (iii) represent consumers and workers, (iv) promote innovation and competition and (v) ensure America's global AI leadership.
The recommendations focus on risk management, including proposals on "disclosure requirements for developers of the most powerful systems, assessing AI's risks for critical infrastructure, and hindering foreign actors' efforts to develop AI for harmful purposes." Specifically, the agencies proposed the use of the Defense Production Act to require that AI developers report and share "vital information" with the Department of Commerce to help identify "large computing clusters"; the production of a draft rule compelling U.S. cloud companies to report if they provide computing power for foreign AI training models in order to potentially stave off "malign activity"; and the requirement of risk assessments "covering AI’s use in every critical infrastructure sector" in order to develop protections for vital services such as the electric grid.
Additional recommendations on innovation include increasing investment in AI and attracting and training workers with AI expertise.
Commentary
Generally, the federal agencies responded to President Biden's Executive Order as expected (see previous commentary). On risk management, the agencies proposed to engage in massive rule making, and for innovation, massive hiring.
If the outcome is an AI rule-writing campaign, the question is whether it can do so without stifling progress, inhibiting free speech, and writing rules that are either wholly ambiguous or impossible to meet. To the extent that the regulators follow up on their current course of prohibiting firms from using AI unless they fully understand the logic behind it, it will be a long time before the benefits of using AI in financial services will be seen. That is a standard that is not possible to meet, as AI is not mechanistic and not entirely predictable in either its results or its "reasoning" (to use a potentially inappropriate and anthropomorphic word).
Related Articles
-
U.S. President Joseph R. Biden issued an Executive Order establishing safety and security standards for artificial intelligence.