High-Risk AI Serious Incident Reporting

Detect, triage, report, investigate and preserve evidence for serious high-risk AI incidents within applicable deadlines.

For a high-risk AI provider, maintain a serious-incident procedure that distinguishes reporting from investigation and correction. Record awareness time, the affected system and the applicable reporting category. Assess the Article 73 timing, any sector-specific reporting limitation and any special route, including the AI Office route where applicable. Report within the relevant rule, using an initial incomplete report where the Act allows rather than waiting for the final investigation. Preserve evidence, investigate and cooperate with authorities. Real-world testing also requires immediate mitigation, or suspension or termination as applicable; check its separate trigger. Do not apply these high-risk-system rules automatically to GPAI model incidents.

On awareness of a serious incident; within the applicable immediate or category-specific legal deadline; follow-up as required.

Incident classification and awareness-time record Reporting deadline and recipient decision Initial and follow-up reports with transmission evidence Preservation, investigation and corrective-action records Authority instructions and cooperation history

Walk through a serious incident with the team and verify the reporting clock, route and information available for an initial report. Check that evidence can be preserved and that investigation or approval steps do not delay required reporting.