AI incident reporting
Current understanding
Emerging U.S. policy area focused on requiring developers or deployers of AI systems to report incidents (harms, failures, misuse) to a federal body, modeled loosely on incident reporting regimes in aviation, cybersecurity, and medical devices. The AI Incident Reporting Act (HR9477) is an early legislative vehicle, introduced and referred to the House Committee on Energy and Commerce; no mandatory federal AI incident reporting regime is in force yet. For developers and platforms this would create new disclosure obligations; for users it would improve visibility into real-world AI harms.
Evidence log
- 2026-06-25 — AI Incident Reporting Act: HR9477 (AI Incident Reporting Act) referred to House Energy and Commerce Committee in June 2026, proposing a mandatory AI incident reporting framework. (novelty: 3)
- 2026-06-25 — AI Flaw Reporting and Security Enhancement Act: 2026 — AI Flaw Reporting and Security Enhancement Act (HR9333) advanced from House committee 35-0, complementing the incident reporting track with a distinct focus on pre-incident vulnerability/flaw disclosure rather than post-incident reporting. (novelty: 3)
- 2026-03-26 — AI Foundation Model Transparency Act of 2026: cross-connection with ai-foundation-model-transparency-act: Both bills route through House Energy and Commerce and would create federal disclosure regimes for AI systems — transparency of model characteristics vs. reporting of operational incidents. (novelty: 3)
- 2026-01-21 — AI OVERWATCH Act: cross-connection with ai-overwatch-act: Both advance federal AI oversight mechanisms in the 119th Congress; potential overlap depending on whether OVERWATCH incorporates incident reporting duties. (novelty: 3)
- 2025-05-15 — AI Whistleblower Protection Act: cross-connection with ai-whistleblower-protection-act: Whistleblower protections complement mandatory AI incident reporting by protecting internal employees who surface AI harms or noncompliance, forming two channels (institutional reporting and individual disclosure) in the emerging AI accountability stack. (novelty: 2)