For Immediate Release
Today the White House Office of Management and Budget (OMB) released its finalized guidance on the use of artificial intelligence by federal agencies. The policy document establishes minimum practices that federal agencies must follow to mitigate the risks of AI to Americans’ safety or rights. It also gives agencies expansive discretion in certain circumstances to waive or otherwise opt out of these practices. While agencies must report these decisions to OMB and generally disclose them to the public, they are not subject to external review.
Faiza Patel, senior director of the Liberty and National Security Program at the Brennan Center for Justice at NYU Law, had the following reaction:
“OMB’s guidance introduces much-needed checks on how federal agencies use AI across a range of government activities, including policing, immigration enforcement, and the delivery of essential public services. These checks serve as a critical line of defense against applications of AI that lead to racial profiling, wrongful arrests, and mass surveillance. But the guidance gives agencies too much leeway to opt out of the safeguards.
“Federal agencies should be exempt from minimum practices only in truly exceptional circumstances, and their decisions to waive or opt out should be subject to rigorous review and challenge by OMB. The stakes are too high, and the potential for harm too great, for the federal government merely to pay lip service to protecting civil liberties and rights as it ramps up deployment of AI.”
Background
The federal government is racing to deploy artificial intelligence technologies across a wide range of functions. In keeping with President Biden’s executive order to advance safe and trustworthy AI, the Office of Management and Budget’s guidance requires federal agencies to adopt practices such as impact assessments, transparency measures, and risk mitigation procedures when they use the technology in ways that potentially impact safety and rights. Agencies may waive these practices in certain cases, including if they deem that compliance would “increase risks to safety or rights overall” or constitute an “unacceptable impediment” to critical agency operations. They can also opt out if they decide that how they use AI does not impact Americans’ rights or safety.
These loopholes give federal agencies, including law enforcement and immigration agencies such as the Department of Homeland Security and the Federal Bureau of Investigation, broad authority to sidestep basic safeguards altogether. Additionally, the OMB guidance does not apply to AI used in national security systems, which will be subject to a separate National Security Memorandum.
Resources
- National Security Carve-Outs Undermine AI Regulations,” (Faiza Patel and Patrick Toomey, Just Security, December 21, 2023)
- Statement of Faiza Patel, AI Insight Forum-National Security (December 6, 2023)
- Comments Submitted to the Office of Management and Budget on Draft Guidance for Government Use of AI (Brennan Center for Justice, December 5, 2023)
