Anthropic cuts its disclosure window for serious bugs to 30 days
High- and critical-severity issues reported from 1 January 2027 get 30 days, others 45; the 7-day window for active exploitation is unchanged, and AI findings without human review are excluded.
- Security & misuse
- Labs & people
- Notable
Anthropic announced changes to its coordinated vulnerability disclosure (CVD) policy on 16 November, cutting the time it gives vendors and maintainers before publishing details from 90 days to 30 for high- and critical-severity issues, and to 45 days for all others. The new deadlines apply to vulnerabilities reported on or after 1 January 2027; reports already in progress keep their existing deadlines.
Two related changes accompany it. Anthropic will aim to escalate to an external vulnerability coordinator after 15 days, rather than 30, if a maintainer does not respond, and will wait 30 days rather than 45 after a patch is available before sharing full technical details. For high- and critical-severity reports more than 90 days old that remain undisclosed, Anthropic said it is giving maintainers 30 more days, after which it will publish them on its dashboard under the updated policy from 1 January.
Anthropic’s stated reason was that disclosure norms predate AI-assisted discovery:
A 90-day window made sense when a vulnerability could be expected to stay private for months. We no longer believe it can, and these changes are intended to keep pace with progress in AI capabilities and strike the right balance for defenders.
— Anthropic, “Updating our coordinated vulnerability disclosure policy timelines”
It said models can now help develop fixes, that a bug one model finds is increasingly likely to be found by another, and that it is seeing a higher rate of rediscovery. The post lists limits. Flaws in hard-to-deploy software such as critical infrastructure, or ones needing a specification change, will get a negotiated timeline. The seven-day window for vulnerabilities under active exploitation is unchanged. The update does not cover AI-generated findings that have had no human review, such as those from its OSS Scanner.
Anthropic acknowledged that shorter deadlines may put more pressure on maintainers. It said it would keep sending patches or mitigations where possible and agree a pace before sending large volumes to one project, and it pointed to its no-cost scanner and a six-month Claude subscription for eligible open-source contributors. Anthropic’s own vulnerability-finding programmes, including Project Glasswing, are the background; the post appeared the same day as Google’s threat group report on attackers automating patch-to-exploit work.
In the commentary
What people were saying around this time — external links, from the record's commentary rail.