BLOG OVERVIEW: On August 27, 2026, OPM Director Scott Kupor issued a memorandum advising agencies to adopt artificial intelligence (AI) across the federal hiring lifecycle. The memo interprets OMB M-25-21 in the hiring context, explaining when an AI use case is "high impact.” Much of the analysis turns on the quality of human review: OPM states that a reviewing official's independent evaluation and adoption of an AI work product can keep the AI output from serving as a principal basis for a decision, but that a nominal review step does not. Agencies must document their determinations to their Chief Artificial Intelligence Officer and continue to meet existing job analysis, veterans' preference, accessibility, and privacy requirements.
On August 27, 2026, the U.S. Office of Personnel Management (OPM) released a memorandum on the use of artificial intelligence (AI) in the federal hiring process. The opening statement notes that agencies may be compromising efficiency and quality in hiring by failing to adopt AI, and points to Executive Order 14170 and the Merit Hiring Plan, both of which push agencies to modernize and speed up federal hiring.
In alignment with those goals, the memo provides guidance on how OMB’s April 2025 Memorandum (M-25-21, Accelerating Federal Use of AI through Innovation, Governance, and Public Trust) applies to hiring, and offers examples of AI uses that OPM generally does not consider “high impact,” despite OMB’s “presumed high impact” categorization. It emphasizes that agencies must evaluate their specific use cases and workflows, with particular emphasis on the degree to which the AI output serves as the principal basis for the decision.
This guidance is noteworthy because while a “high impact” designation does not preclude an agency’s use of an AI tool, it triggers significant additional risk-management requirements that may make agencies more hesitant to adopt AI in their hiring process. OMB's high-impact standard, OPM’s guidance for how human review factors into that determination, the requirements that apply regardless of an AI use's impact categorization, and OPM's recommended next steps for agencies are analyzed below.
Under OMB’s M-25-21, an AI use is "high impact" when its “output serves as a principal basis for decisions with a legal, material, binding, or significant effect on” interests that include access to employment. Certain categories are “presumed high impact” by OMB, including pre-employment screening, hiring and termination, reasonable accommodation, performance management, pay or promotion, recommending disciplinary action, and reassignment of workers to new tasks or teams.
If an agency concludes that a presumed high-impact use isn't actually high impact, an official documents that decision to the agency’s Chief Artificial Intelligence Officer (CAIO). Uses that remain high impact need either to meet the risk-management practices in M-25-21 (e.g., pre-deployment testing, an AI Impact Assessment, ongoing monitoring, human oversight, and other risk-management measures) or complete a waiver (Appendix B of the memo).
For an AI use to generally fall outside the high-impact category, the key question is whether the AI output serves as the principal basis for the decision. OPM’s guidance focuses on whether a qualified official independently reviews the underlying information and makes the decision based on that review. It further states that a formal human sign-off is not sufficient; the record should demonstrate that the official considered the underlying applicant information and used the AI as an assistive tool, rather than simply adopting its recommendation or score.
This distinction is particularly important for AI uses that OMB presumes to be high impact, such as pre-employment screening. OPM’s Appendix A provides several examples of when these uses may nevertheless fall outside the high-impact definition1. For example, AI-assisted résumé screening may be non-high-impact when the AI identifies or summarizes information from an applicant’s record, the official independently reviews the underlying documentation, and the decision is based on that documentation rather than the AI-generated summary. If the official simply approves the AI’s recommendation, the use remains high-impact.
OPM is careful to note that a non-high-impact determination does not relax other federal agency requirements. Agencies must still comply with job analysis and validation rules, veterans’ preference, accessibility and reasonable-accommodation requirements, Privacy Act requirements, and applicable recordkeeping and reconsideration procedures.
OPM concludes by encouraging agencies to loop in their privacy, information technology, and data officers, and, while it continues to expand AI capabilities within the USA Suite of hiring tools, tells agencies not to wait to get started on their own efforts.
DCI will continue to monitor developments in AI and federal hiring policy.
1 OPM notes the assessments in the memo and in Appendix A reflect OPM’s views on how OMB’s M-25-21 applies to common Federal hiring use cases and were developed in consultation with OMB, but they are not official OMB determinations.