Trending: On-device modelsSearch
iHeartGeek
iTECH

Anthropic shortens vulnerability disclosure windows as AI finds flaws faster

Anthropic will disclose serious vulnerabilities faster from 2027, citing AI-assisted discovery and a rising risk of independent rediscovery.

A computer screen displaying coloured source code, illustrating software vulnerability research

Anthropic is tightening the timetable for publicly disclosing software vulnerabilities, saying AI-assisted discovery is making it more likely that flaws will be found again before a traditional remediation window expires. The company announced the change in an update to its coordinated vulnerability disclosure policy on November 16.1

For vulnerabilities reported to vendors or maintainers from January 1, 2027, Anthropic says high- and critical-severity issues will move from a 90-day disclosure window to 30 days. Other vulnerabilities will receive 45 days. Reports already in progress on that date will keep their existing deadlines.1

The policy also changes what happens when maintainers do not respond. Anthropic says it will aim to involve an external vulnerability coordinator after 15 days, rather than 30. Once a patch is available, it plans to wait 30 days before publishing full technical details, down from 45 days.1

The seven-day deadline for vulnerabilities under active exploitation is unchanged. Anthropic also says the revised policy will not apply to AI-discovered findings that have not received human review, including findings from its OSS Scanner service.1

Anthropic frames the shift as a response to faster AI-enabled vulnerability discovery and a higher rate of independent rediscovery. Its argument is straightforward: a flaw that one model can find may not remain private for three months if another model can uncover it too. That does not make every bug easy to fix, though. The company acknowledges that critical infrastructure and standards-related flaws may need longer remediation timelines, and says it will work with maintainers on exceptions.1

The announcement is a policy change, not evidence that shorter deadlines will improve security outcomes in practice. The immediate effect will be greater pressure on maintainers to confirm reports, ship patches or negotiate a remediation plan more quickly. It also gives security teams a clearer signal that a report left unanswered may reach an outside coordinator sooner.

Our opinion

Anthropic is right to revisit disclosure rules as automated vulnerability research improves, but a shorter clock should not become a blunt measure of responsible security work. Thirty days may be sensible for a well-understood web flaw and unrealistic for a defect buried in infrastructure that takes months to test safely. The useful part of this policy is the explicit room for exceptions and human review. The important test will be whether Anthropic applies that flexibility consistently, rather than treating speed as the only measure of a successful disclosure.

For now, the change is best understood as a warning to maintainers: the private period around a serious vulnerability is becoming harder to protect. Teams that still depend on a 90-day assumption should check how quickly they can triage, communicate and publish a fix before the new timelines arrive.