ToolNavs Find Useful AI Tools
Submit Sign in
Back to AI information
Anthropic Cuts Vulnerability Disclosure Window to 30 Days: AI Finds Bugs Too Fast for 90 Days

Anthropic Cuts Vulnerability Disclosure Window to 30 Days: AI Finds Bugs Too Fast for 90 Days

AI information • Admin • • 17 views

On November 16, Anthropic updated its coordinated vulnerability disclosure (CVD) policy: the disclosure window for high- and critical-severity vulnerabilities drops from 90 days to 30 days, and to 45 days for everything else. The new timelines apply to vulnerabilities reported to vendors on or after January 1, 2027; reports already in progress keep their existing deadlines. Anthropic's reasoning is blunt: frontier models now find vulnerabilities so fast that a 90-day window no longer makes sense.

Two further changes come with the shorter window. If a vendor does not respond to a report within 15 days, down from 30, Anthropic will escalate it to an external vulnerability coordinator; and once a patch is available, the wait before publishing full technical details shrinks from 45 days to 30. The seven-day disclosure window for vulnerabilities under active exploitation stays unchanged. There is also a backlog to clear: high- and critical-severity vulnerabilities that Anthropic reported more than 90 days ago and that still have not been disclosed publicly will get their maintainers one more 30-day notice, after which the reports will be published on Anthropic's vulnerability dashboard starting January 1, 2027.

Open-source maintainers will feel the squeeze most. With the deadline cut by two-thirds, the whole cycle of patching, testing and shipping has to move faster. Anthropic acknowledges the pressure and promises to keep attaching actionable guidance to every report, with a patch or mitigation wherever possible, and to agree on a sustainable pace with maintainers before sending a large volume of findings to a single project. Exceptions remain for flaws that cannot be fixed quickly: software that is hard to deploy, such as critical infrastructure, and flaws that require changes to a specification or RFC will get a negotiated remediation timeline. The new policy covers only findings that have received human review; raw output from AI scanning is excluded.

For ordinary users, the effect is invisible but real: a shorter path from discovery to disclosure also compresses the patching window, so keeping software updated matters more. January 1, 2027 is the dividing line between the old and new rules, and vulnerabilities reported before then stay on the old timelines. For maintainers, Anthropic is offering some cushion: the free OSS Scanner for regular code scans, and the Claude for Open Source Program, which gives eligible maintainers six months of Claude Ultra at no cost.

Recommended Tools

More