Sulat.com / Writing on AITuesday, November 17, 2026
Artificial Intelligence/17 November 2026

Anthropic cuts its bug disclosure deadline from 90 days to 30 for high and critical flaws

Anthropic's Frontier Red Team says reports sent from 1 January 2027 get 30 days for high and critical bugs and 45 for the rest. Chris Rohlf says it will still send patches and agree a pace with maintainers.

Glossy 3D Anthropic logo in clay-orange glass beside an hourglass nearly out of sand and a torn calendar block.

At a glance

  • Anthropic's Frontier Red Team says its disclosure deadline drops from 90 days to 30 for high and critical bugs and to 45 days for all others, for reports sent on or after 1 January 2027.
  • Anthropic will aim to escalate to an external coordinator after 15 days of maintainer silence instead of 30, and will wait 30 days after a patch, down from 45, before sharing full technical details.
  • Anthropic says AI has made fixes faster and rediscovery of the same bug more likely, and that it no longer believes a bug can be expected to stay private for months.
  • The seven-day window for bugs under active exploitation stays, and the update doesn't apply to AI-discovered findings that no human has reviewed, such as those found by OSS Scanner.
  • Chris Rohlf says the shorter timelines "may put more pressure on OSS maintainers and vendors", and that Anthropic will keep sending a patch or mitigation where possible and agree a sustainable pace first.

Anthropic is cutting the disclosure deadline in its coordinated vulnerability disclosure (CVD) policy from 90 days to 30 for high- and critical-severity bugs, and to 45 days for everything else. The company’s Frontier Red Team announced it in a post dated 16 November, and the new deadlines apply to bugs reported to vendors and maintainers on or after 1 January 2027.

Chris Rohlf, a security engineer on Anthropic’s Frontier Red Team, announced the change on X. In coordinated disclosure, the finder tells the software’s maker privately and publishes the details once a deadline passes. Rohlf says Anthropic is making the change to “keep pace with the rapid improvement of AI driven vulnerability discovery capabilities”.

Before and after: 90 days becomes 30 days for high and critical bugs and 45 days for all other bugs, for reports from 1 January 2027.
High and critical bugs drop from 90 days to 30, and every other bug to 45.

What changes on 1 January

Anthropic’s post lists two more changes alongside the shorter deadlines. If a maintainer doesn’t answer the first report, Anthropic will now aim to escalate to an external vulnerability coordinator after 15 days instead of 30. After a patch ships, it will wait 30 days instead of 45 before sharing full technical details. Reports already in progress on 1 January keep their existing deadlines.

ItemBeforeFrom 1 January 2027
Deadline, high and critical bugs90 days30 days
Deadline, all other bugs90 days45 days
Escalate if the maintainer is silent30 days15 days
Wait after a patch before full details45 days30 days
Bug under active exploitation7 days7 days
Sources: Anthropic’s November 2026 post and its disclosure policy page, last updated 6 March 2026.
Timeline: report on day 0, escalation at day 15, deadlines at day 30 for high and critical bugs and day 45 for others, then 30 days after a patch.
A silent maintainer now gets 15 days, not 30, before Anthropic escalates.

Some bugs are already past the old deadline. Anthropic says it reported high- and critical-severity vulnerabilities more than 90 days ago that haven’t been disclosed yet, and it’s telling those maintainers they get 30 more days. After that, the post says, Anthropic will publish the reports on its CVD dashboard in line with the updated policy, starting 1 January 2027.

Why Anthropic shortened it

Anthropic says disclosure norms, its own included, were written before AI could find vulnerabilities. They assumed fixes moved at human speed and that nobody else would independently rediscover the same bug. The post says neither holds now: models can help develop fixes, and a bug one model finds is “increasingly likely to be found by another”. Anthropic says it’s seeing a higher rate of rediscovery, and that a 90-day window made sense when a bug could be expected to stay private for months.

The volume is large. In its 8 October post launching OSS Scanner, Anthropic said it had found more than 29,000 candidate vulnerabilities over six months and manually reviewed about 6,000 of them. Google’s Project Zero made a similar move on 13 November, replacing its 90+30 policy with a fixed deadline that drops from 90 days to 30 by January 2028, and it also pointed to large language models.

What maintainers get

Anthropic’s post acknowledges that shorter timelines “may put more pressure on software maintainers”. It says it will keep including actionable guidance with each report, along with a patch or mitigation where possible, and will keep agreeing a sustainable pace with maintainers before sending a large volume of findings to one project. Flaws in hard-to-deploy software, such as critical infrastructure, and flaws that need a change to a specification or RFC can’t always be fixed in 30 or 45 days. For those, Anthropic says it will work with the vendor on a remediation timeline.

Checklist: 7 days for actively exploited bugs, a fix with each report, agreed pace, and AI findings without human review not covered.
The 7-day rule for bugs under attack stays, and unreviewed AI findings sit outside the change.

The seven-day window for any vulnerability under active exploitation stays as it is. The update also doesn’t apply to AI-discovered findings that haven’t had human review, such as those found by OSS Scanner, a point Rohlf repeats in his post.

Anthropic points maintainers to three programs. Maintainers of critical open-source projects can apply to OSS Scanner, a no-cost service that periodically scans their code with Anthropic’s strongest models and sends them findings directly. Eligible maintainers and contributors can apply to the Claude for Open Source Program for six months of Claude Ultra at no cost. Security professionals, including open-source maintainers, can apply to the Cyber Verification Program, which tailors Claude’s safeguards for legitimate defensive work.

Anthropic’s disclosure policy page still shows the 90-day default and a “Last updated Mar 6, 2026” line. The new deadlines start applying to reports sent from 1 January 2027.

Sources


Discover more from Sulat AI & Tech News

Subscribe to get the latest posts sent to your email.