Anthropic cuts its bug disclosure deadline to 30 days, saying another AI will find the bug
Its own dashboard shows maintainers had patched 629 of the 5,087 bugs it reported, and now they get a third of the time.
Anthropic is cutting the time software makers get to fix a serious bug before it goes public. In a post on its research site, the company says its disclosure window for high- and critical-severity flaws drops from 90 days to 30, and to 45 days for everything else.
The new deadlines apply to bugs it reports on or after January 1, 2027. Anthropic gives one reason, and it's about AI itself: a bug one model can find is now more and more likely to be found by another.
- Who
- Anthropic
- High and critical bugs
- 90 days down to 30
- Everything else
- Down to 45 days
- Starts
- Reports made on or after January 1, 2027
- Under active exploitation
- Seven days, unchanged
What changes
A disclosure window is how long the finder keeps quiet while the fix gets made. Anthropic's current policy page, last updated March 6, 2026, calls 90 days the industry standard and allows a 14-day extension when a maintainer is close to a fix.
Two smaller deadlines move with the main one. If a maintainer doesn't answer its first report, Anthropic will go to an outside vulnerability coordinator after 15 days instead of 30. And once a patch is out, it'll wait 30 days instead of 45 before publishing full technical details.
A 90-day window made sense when a vulnerability could be expected to stay private for months. We no longer believe it can
There's a backlog, too. Some high- and critical-severity bugs Anthropic reported more than 90 days ago are past the current deadline and still aren't public. The company says it's giving those maintainers another 30 days and will then publish the reports on its dashboard.
Who has to move faster
The people on the other end of these reports are mostly open-source maintainers. Anthropic's own disclosure dashboard shows the size of the gap they're already working with.
As of November 2, 2026, Anthropic says it had disclosed 5,087 vulnerabilities across 680 open-source projects, and that 629 of them had been patched as far as it knows. That's about one in eight. Some fixes ship without an advisory, so the real figure may be higher, but the dashboard can't show it.
That's what this means for anyone running software: a third of the old window to patch whatever these models turn up, or see the details published anyway. I'd read it next to what Sam Altman told Stanford students about patching, which we covered in October.
What's left out
Anthropic does carve out exceptions. Critical infrastructure and flaws that need a change to a standard or an RFC can't always be fixed in 30 or 45 days, and for those it says it'll agree a timeline with the vendor. Findings no human has reviewed, such as those from its OSS Scanner, aren't covered at all.
For maintainers, the offer is help with the work. Anthropic says each report will keep coming with guidance and, where it can, a patch or mitigation. It also points to free programs, including six months of Claude Ultra for eligible open-source maintainers. The new clock starts on January 1, 2027.
More on Anthropic
- Anthropic says Mythos got Claude running on AMD chips in three days, almost unaidedNovember 12, 2026
- Anthropic launches Claude Sheets and Whiteboards, but Claude Meetings is still coming soonNovember 10, 2026
- AI Futures Project now expects a superhuman AI researcher in 2027, citing METR's Anthropic reportNovember 10, 2026
- Anthropic starts its IPO roadshow, but a single founders' share will carry 50.1% of the voteNovember 10, 2026