Anthropic is shortening the disclosure window in its coordinated vulnerability disclosure policy from 90 days to 30 days for high- and critical-severity issues, and to 45 days for all other issues. In a 16 November post by its Frontier Red Team, the company said the new timelines will apply to vulnerabilities reported to vendors and maintainers on or after 1 January 2027. Reports already in progress on that date will keep their existing deadlines.
Two related changes accompany the shorter window. Anthropic said it will aim to escalate to an external vulnerability coordinator after 15 days, rather than 30, if a maintainer does not respond to the initial report. It will also wait 30 days after a patch is available, rather than 45, before sharing full technical details. “The rest of our policy remains the same.”
The company also said some high- and critical-severity vulnerabilities that it reported more than 90 days ago, and that now exceed the deadline in the current policy, have not yet been disclosed. It is notifying those maintainers that they have an additional 30 days, after which it will publish the reports on its dashboard under the updated policy, starting on 1 January 2027.
Why Anthropic says the 90-day window no longer holds
Anthropic argued that many coordinated-disclosure norms were written before AI systems could discover vulnerabilities at scale. Those policies, it said, assumed that fixes moved at human speed and that independent rediscovery of the same bug was unlikely. “Neither assumption holds today.” Models can now help develop fixes, and a bug that one model can find is “increasingly likely to be found by another.” The company said it is seeing a higher rate of rediscovery as those capabilities increase.
“A 90-day window made sense when a vulnerability could be expected to stay private for months. We no longer believe it can.” The changes, it said, are intended to keep pace with progress in AI capabilities and to strike “the right balance for defenders.”
Some flaws still cannot be fixed quickly, including problems in hard-to-deploy software such as critical infrastructure, and flaws that require changes to a specification or RFC. In those cases, Anthropic said, it will work with the vendor on a remediation timeline. Its seven-day disclosure window for any vulnerability under active exploitation is unchanged. The new policy will not apply to AI-discovered findings that have not received human review, such as those found by OSS Scanner.
The company acknowledged that shorter timelines “may put more pressure on software maintainers.” It said it will continue to include actionable guidance, and a patch or mitigation where possible, and will agree a sustainable pace with maintainers before sending a large volume of findings to a single project. It pointed maintainers to OSS Scanner, six months of Claude Ultra through the Claude for Open Source Programme, and its Cyber Verification Programme.
The policy shift sits in a wider debate about how quickly AI-found bugs should be disclosed, including Bitcoin Core’s decision to end waits of up to a year for bugs that language models can find easily, Anthropic’s earlier Cyber Mission and free open-source scanner, and Trend Micro’s Zero Day Initiative sending unpatched Linux kernel flaws to public lists. curl’s move to take low- and medium-severity reports in public and Anthropic’s earlier account of unreleased-model Chrome findings are neighbouring disclosure stories, not this policy change. Google’s Project Zero, in a 13 November post, has separately set out its own shorter deadlines, moving to 60 days from 2 February 2027 and to 30 days from 3 January 2028. That is background on the same industry argument, not Anthropic’s rule.
A shorter window from 2027 is not an immediate cut-off
The new 30- and 45-day clocks do not start today. They apply to vulnerabilities reported on or after 1 January 2027. Anything already in the queue keeps the deadline it already had. Reading the announcement as an immediate cut from 90 days would ignore the date Anthropic actually set, and would also ignore the extra 30 days it is giving maintainers of high- and critical-severity reports that are already overdue under the current policy.
The unchanged seven-day window for active exploitation, and the promise to agree a timeline for flaws in critical infrastructure or in specifications, are also part of the official record. So is the exclusion of unreviewed OSS Scanner findings. A shorter default is not a claim that every bug will be fixed, or disclosed, in 30 days.
Anthropic’s reason for the change is its own: it no longer believes a 90-day private window will hold, because models can rediscover the same bugs and can help write fixes. That is a policy judgment, not a published measurement of how often rediscovery is now occurring, or of how often maintainers will meet the new dates. The next documented test will be whether the 1 January 2027 reports, and the overdue cases now being given 30 more days, are closed or published on the timetable the company has set.
Sources
- Anthropic Frontier Red Team, “Updating our coordinated vulnerability disclosure policy timelines”, 16 November 2026
- Google Project Zero, policy update, 13 November 2026 (background only)
Get TSN’s biggest tech stories, explained.
News and insights across AI, tech, crypto and cybersecurity. Every source linked. Unsubscribe anytime.
Get the newsletterRelated stories
- Bitcoin Core Ends Waits of Up to a Year for Bugs LLMs Can Easily Find is the closest recent TSN account of a project shortening disclosure timelines because language models can find the same bugs more quickly, and the right next stop for that debate.
- Anthropic Launches the Anthropic Cyber Mission: Critical-Infrastructure Programme and Free Open-Source Scanner
- Trend Micro’s ZDI Sent Unpatched Linux Kernel Flaws to Public Lists
Follow @tsnmediahub on X for breaking AI news as it happens.
Follow on X





