Eigen RadarAI
Analysis

Anthropic cuts severe-flaw disclosure deadlines to 30 days from January

Anthropic will shorten the time between privately reporting a serious software vulnerability and publicly disclosing it. Reports sent from January will have a 30-day deadline for high and critical severity, down from 90 days. Other findings will get 45 days. The company is also accelerating escalation when maintainers do not respond, while retaining exceptions for difficult repairs.

Artificial Intelligence··Evening
High-angle overhead view of three security reviewers standing around a circular table with closed red and amber dossier folders in a sunlit San Francisco AI office atrium.

New reports receive shorter deadlines from January

AI company Anthropic will reduce its coordinated vulnerability disclosure window from 90 days to 30 for high- and critical-severity flaws. Other issues will receive 45 days. The policy governs the time between notifying a software vendor or maintainer privately and making the finding public. It applies to reports sent on or after 1 January; reports already in progress retain their existing deadlines.[1], [2]

Unanswered notifications will be escalated sooner

When a maintainer does not respond to the initial report, Anthropic will aim to involve an external vulnerability coordinator after 15 days, instead of 30. The wait between a patch becoming available and release of full technical details will fall from 45 to 30 days. The seven-day window for vulnerabilities under active exploitation remains unchanged.[1]

For serious flaws reported more than 90 days ago that remain undisclosed, the company is notifying maintainers of an additional 30 days before publication on its dashboard. Anthropic says independent rediscovery of the same vulnerability by different models is becoming more common.[1]

Difficult repairs can have an agreed timetable

Critical-infrastructure software and problems requiring changes to a technical specification or internet standard can need longer repairs. Anthropic will work with vendors on a remediation timetable in those cases. Findings discovered by AI but not reviewed by a human are excluded from this update. The company will continue supplying actionable guidance and, where possible, patches or mitigations. Before sending large batches of findings to one project, it will retain agreement with maintainers on a sustainable pace.[1]

References

  1. News sourceAnthropicAnthropic will cut disclosure windows for severe flaws to 30 days in January↩1↩2↩3↩4
  2. News sourceHelp Net SecurityAnthropic vulnerability disclosure deadline drops to 30 days for serious bugs↩