Policy Briefing — EU AI Act serious-incident reporting consultation closes
The Commission’s Article 73 consultation on serious-AI-incident reporting closes 7 November 2025, pairing draft guidance and templates so high-risk AI providers can operationalise mandatory submissions before the EU AI Act applies in August 2026.
Executive briefing: The European Commission issued draft guidance and a reporting template for serious incidents under Article 73 of the EU Artificial Intelligence Act, opening a consultation that closes on 7 November 2025. Providers of high-risk AI systems must use the window to align governance, service desks, and evidence lockers with the regulation’s requirement to notify market-surveillance authorities within 15 days—or faster for widespread or fatal events—once a causal link is established.
What changed
- Draft guidance released. The consultation package clarifies definitions, cross-regime interoperability, and expects providers to rehearse how incident data flows to national authorities and the EU AI Office using the standard template.
- Hard timelines. Article 73(2)-(4) and (7) set 15-day, two-day, and 10-day escalation thresholds depending on severity, and require Commission guidance by 2 August 2025—leaving this November consultation as the final feedback checkpoint.
- Template dependencies. Providers must document detection triggers, corrective actions, and coordination with notified bodies so authorities can respond within seven days as mandated by Article 73(8).
Action items
- Map incident owners. Confirm 24/7 accountable contacts, escalation chains, and tooling integrations (ticketing, logging, call trees) for every high-risk AI system, including deployer-operated variants.
- Dry-run the template. Populate the Commission’s draft form with recent near-miss events to validate data availability, evidence retention, and interface with broader safety management systems.
- Align with sector regimes. Coordinate with existing safety or supervisory reporting (medical devices, financial services, critical infrastructure) so Article 73 submissions don’t duplicate or contradict domain rules.
Enablement moves
- Track the consultation’s final guidance update and refresh playbooks ahead of the AI Act’s August 2026 applicability, keeping board dashboards synced with Article 73 metrics.
- Integrate lessons with OECD AI incident frameworks to support cross-border transparency commitments referenced by the Commission.