OPM memo outlines AI use in federal hiring, sets risk management for high-impact tools

The memo, released Thursday, directs agencies to review AI tools used in hiring and classify them as either 'high-impact' or 'presumed high-impact' based on whether the AI output serves as a principal basis for decisions with legal, material, or significant effects on civil rights, privacy, access to government programs, critical infrastructure, or public safety.

For 'high-impact' cases, agencies must implement minimum risk management practices or request a waiver from their chief AI officer. Agencies must also document and justify any determination that a 'presumed high-impact' tool does not qualify as high-impact, including controls such as human review, sampling, audit logs, and performance monitoring.

The memo identifies 'presumed high-impact' cases as tools that draft job descriptions, qualification requirements, evaluation statements, interview questions, screen resumes, review documents to determine veterans’ eligibility, or score candidates. However, OPM notes these tools are generally not high-impact if a human independently reviews a candidate’s record or other safeguards are in place.

OPM’s chief information officer, Adam Starr, stated in a blog post that the current federal hiring process, which averages 87 days, is not working and that AI is an 'obvious step' to speed up hiring while improving the process for job applicants and taxpayers.

In addition to the memo, OPM is exploring AI features for USAJOBS, USA Staffing, and USA Hire, including improved search capabilities, plain language job titles and descriptions, and other tools to help job seekers connect with federal openings.