Shortcomings: Anthropic Speeds Up, with Exceptions

Anthropic plans to give developers less time before publicly disclosing reported vulnerabilities. Negotiated timelines will still be possible for certain complex fixes, while automated discoveries without human validation will remain excluded from this policy.
Anthropic Maintains Negotiable Timelines
Fixing a vulnerability affecting critical infrastructure or a technical standard may require more than 30 or 45 days. Anthropic intends to agree on a deadline with the relevant vendor. For an already exploited flaw, the timeline remains seven days. Discoveries produced by AI without human verification, particularly through OSS Scanner, do not fall under this policy.
The company acknowledges that a tighter timeline may increase the pressure on maintainers. It promises to continue providing recommendations for fixes with each report and, when possible, a patch.
New Reports Will Transition in January 2027
Announced on Monday by Frontier Red, the Anthropic team, the new rule will apply to reports made starting January 1, 2027. For vulnerabilities classified as high or critical, disclosure will occur after 30 days, instead of the 90 days previously granted. Others will have a 45-day deadline. If there is no response to the initial report, an external coordinator will be contacted after 15 days, down from 30 currently.
Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisissez votre rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
The dissemination of all technical details will follow the release of the patch after a 30-day wait, compared to 45 days before. Open cases will retain their previous deadlines. However, some serious flaws reported for more than 90 days have exceeded the current deadline without being made public: Anthropic states it will inform their maintainers that they have an additional 30 days. It plans to then publish them on its dashboard, in accordance with the new policy, starting January 1, 2027.
Google Also Plans to Shorten Its Deadlines
Anthropic justifies this change by the increased speed with which, it claims, models detect bugs. The company believes that the previous 90-day deadline was appropriate when a vulnerability could remain confidential for several months, a hypothesis it no longer holds. It explains that the previous rules assumed fixes at a human pace and that the same defect was rarely discovered by two different people. According to it, a bug identified by a model now has an increasing likelihood of being found by another.
At Google, the Project Zero team also plans to shorten the waiting period before disclosure. Its timeline will be reduced to 60 days after reporting starting February 2, 2027, and then to 30 days starting January 3, 2028.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.