EU AI Act changes extend deadlines for high-risk HR tools
The EU Digital Omnibus has postponed AI Act obligations for high-risk recruitment and career-management systems until December 2, 2027, and for AI built into regulated products until August 2, 2028. Prohibitions and AI literacy obligations have applied since February 2025, while transparency rules have applied since August 2026. Employers should use the additional time to inventory tools, strengthen governance, address bias, train staff and define permitted recruitment uses.
Inventories should cover approved software, tools adopted independently by local teams and AI features added to existing HR platforms. Candidate ranking, promotion recommendations and employee profiling are high risk, while narrow administrative functions may qualify for exceptions that require documented assessments. Workplace emotion recognition is prohibited except for medical or safety purposes. Employers should verify vendor compliance, secure contractual access to documentation and logs, and give human supervisors genuine authority to challenge recommendations or suspend systems.
Bias controls should examine outcomes across groups, discriminatory proxies and recruitment prompts, with testing repeated after significant changes. Training should reflect employees’ responsibilities and leave evidence of the measures taken. Recruitment uses also require clear purposes, appropriate GDPR assessments and safeguards for affected candidates. Using a general-purpose chatbot to rank applicants could expose an employer to provider obligations. Compliance processes should preserve meaningful human review, explanations and routes to challenge employment decisions.