OMB Finalizes Long-Awaited Guidance on Federal AI Use

Today, the Office of Management and Budget (OMB) released its final guidance on AI governance, innovation, and risk management within federal agencies. The guidance follows a draft memorandum released last November and stems from obligations set out in President Biden’s Executive Order 14110 on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence.

While the true impact of OMB’s final guidance will depend on federal agencies’ implementation, the guidance advances critical AI risk management practices championed by EPIC, including (1) pre-deployment AI impact assessments, (2) transparency and notification obligations around AI systems in use and their negative impacts on individuals, and (3) additional AI procurement obligations. After soliciting and incorporating public comments on the draft memorandum, OMB also updated its final guidance in several important ways, including:

1. Requiring additional transparency and reporting around agencies’ AI processes and systems in use, including additional information on waivers from required minimum AI risk management practices and disclosure of agencies’ custom AI code, models, and model weights.

• Recommending, but not requiring, additional risk management practices for the procurement and use of biometric identification systems, which EPIC has long pushed to ban.

• Requiring agencies to provide prominent and readily accessible mechanisms for individuals to opt-out of AI decision-making in favor of human decision-makers, which agencies must ensure are fair and nondiscriminatory.

• Expanding required agency processes for consulting and incorporating feedback from impacted communities, civil society organizations, and the public at large—as well as requiring agencies to pause AI use if consultations suggest an AI use case will cause more harm than good.

OMB’s final guidance goes further than most U.S. AI regulations, but EPIC stills sees room for serious improvement, such as (1) expanding the scope of rights- and safety-impacting AI designations beyond systems that serve as the “principal basis” for a decision, (2) adding stricter oversight for agency waivers from OMB requirements and agency decisions around rights and safety-impacting AI systems, and (3) greater oversight over national security and military AI applications. EPIC will continue to vigorously pursue state and federal AI regulations to ensure responsible, rights-preserving, and privacy-protective AI use across the country.