Are frontier AI labs keeping the dated safety promises they made?

29 lab promises · 6 regulatory · updated 2026-08-05 · GitHub · Open data (JSON) · RSS

⚠ Overdue right now · 2

OpenAI governance ⚠ contested ⚠ under review Overdue

Review the Preparedness Framework at least once a year

OpenAI’s Preparedness Framework v2 (2025-04-15) commits to reviewing and potentially updating the framework at least once a year; the next annual review fell due around 2026-04-15.

112 days overdue
committed 2025-04-15 due 2026-04-15 · now
Sources: OpenAI (Preparedness Framework v2) ↗ archived · OpenAI (context) ↗ archived · auto-checked daily · reviewed 2026-06-18
Committed text
“review and potentially update the Preparedness Framework for continued sufficiency at least once a year”
Why this ruling

The 2026-04-15 deadline is derived from the v2 "at least once a year" annual-review cadence (Preparedness Framework v2, 2025-04-15), not an OpenAI-stated date. No 2026 Preparedness Framework review had been published as of 2026-06-18. The May 2026 Frontier Governance Framework is a separate document and does not update the Preparedness Framework.

Anthropic safety-framework ⚠ contested ⚠ under review Overdue

Define ASL-4 safeguards before reaching ASL-3

Anthropic’s Responsible Scaling Policy v1.0 stated its commitment was to write the ASL-4 measures before any model reaches ASL-3 capabilities.

Trigger fired 2025-05-22 — ruling due
Sources: Anthropic (RSP v1.0) ↗ archived · Anthropic (context) ↗ archived · auto-checked daily · reviewed 2026-06-18
Committed text
“define ASL-4 evaluations before we first train ASL-3 models”
Why this ruling

The original v1.0 trigger is documented. The trigger has since elapsed — Anthropic activated ASL-3 with Claude Opus 4 on 2025-05-22 — and v3.0 (Feb 2026) restructured away from the ASL-4 framing; whether the loosely-specified ASL-4 commitment was satisfied is genuinely disputed.

Upcoming regulatory milestones

Context, not scored. These are laws (e.g. the EU AI Act), not promises a lab made — shown as countdowns, never marked "missed".

Multi-labgovernance

EU General-Purpose AI Code of Practice published

The European Commission published the final General-Purpose AI Code of Practice on 2025-07-10. It is a voluntary instrument under the EU AI Act, not a promise a lab made.

in force since 2025-07-10
Source: European Commission ↗
Multi-labgovernance

EU AI Act general-purpose AI obligations begin to apply

Under EU AI Act Article 113, the general-purpose AI model obligations (Chapter V) began to apply on 2025-08-02, 12 months after the Act entered into force.

in force since 2025-08-02
Source: EU AI Act ↗
Multi-labgovernance

EU AI Act obligations apply to GPAI models already on the market

Under EU AI Act Article 111, providers of general-purpose AI models placed on the market before 2025-08-02 must comply with the GPAI obligations by 2027-08-02.

in 362 days
Source: EU AI Act ↗
Multi-labgovernance

EU AI Act high-risk (Annex III) obligations apply

The EU AI Act originally applied Annex III (standalone, use-based) high-risk obligations from 2026-08-02. The Digital Omnibus deal (provisional political agreement, May 2026) defers them to 2027-12-02.

in 484 days
Source: Hogan Lovells ↗
Multi-labgovernance

EU AI Act high-risk (Annex I) obligations apply

The Article 6(1) obligations for high-risk systems that are safety components of products covered by EU product law (Annex I) originally applied from 2027-08-02. The Digital Omnibus deal (provisional agreement, May 2026) defers them to 2028-08-02.

in 728 days
Source: Hogan Lovells ↗
Multi-labgovernance

EU AI Act first Commission evaluation and review

Under EU AI Act Article 112, the Commission must evaluate the need to amend the high-risk list and related provisions by 2028-08-02, then every four years.

in 728 days
Source: EU AI Act ↗

Latest updates

All updates · RSS

Hold them to it.

One email when a tracked deadline flips overdue, or a ruling changes. No spam.