The White House’s Office of Management and Budget (OMB) has officially set a Dec. 16 deadline for Federal agencies to submit annual inventories of their AI use cases.
The finalized AI reporting guidance – distributed to agencies Aug. 14 but made public Friday – follows draft guidance released by OMB in March alongside its governmentwide AI policy in line with the Biden-Harris administration’s AI executive order (EO).
The new guidance is the latest iteration of how Federal agencies will go about collecting and producing lists of their planned, new, and existing AI use cases.
The inventories were first established under a 2020 Trump-era EO on AI that have now have been expanded upon by the Biden administration.
By Dec. 16, 2024, agencies – apart from the Defense Department and intelligence community – are required to disclose the information about their AI uses to OMB via an online form and also post a “machine-readable CSV of all publicly releasable use cases” on their website.
The inventories will include more standardized categories and multiple-choice responses for agencies under OMB’s new guidance.
For all use cases, agencies must report basic summary information, including the AI’s intended purpose, expected benefits, and outputs. For a subset of AI use cases, OMB is requiring agencies to report on additional information regarding development, data and code, and AI enablement and infrastructure.
The finalized guidance also requires agencies to review each of their current and planned uses of AI to determine whether it matches the definition of safety-impacting or rights-impacting. Agency chief AI officers (CAIOs) must make this full list publicly available on their websites by December.
For only those use cases that have been identified as rights-impacting or safety-impacting, agencies are required to report on information regarding the AI’s compliance with OMB’s risk management practices. This includes whether there’s a process for testing and monitoring the AI’s performance, whether groups impacted by the AI have been consulted in its development, and how the agency is mitigating discrimination from the use of AI, among others.
For a particular use of AI that cannot feasibly meet OMB’s minimum risk management practices by December, an agency may request an extension by Oct. 15 for up to one year, OMB said.
The guidance also states that an agency CAIO may waive one or more of OMB’s required minimum risk management practices. A waiver must provide detailed justification, based upon a system-specific and context-specific risk assessment, that fulfilling the requirement would increase risks to safety or rights overall or would create an unacceptable impediment to critical agency operations.
CAIOs must make a summary of each waiver and its corresponding justification publicly available on their websites.
