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, the company said in a post on 16 November. By our arithmetic, the window for the most serious bugs becomes three times shorter.
The new timelines apply to vulnerabilities reported to vendors and maintainers on or after 1 January 2027. Anthropic gives AI as the reason: “the speed at which AI is improving has changed how we think about our own disclosure practices”.
Business Pill · Least Privilege
A short Business Engineer explainer on giving software and AI agents only the access a task needs. It teaches the general security idea only and does not discuss Anthropic’s disclosure policy.
The key insight: As we read the post, Anthropic is pricing in rediscovery: it says a bug one model can find is increasingly likely to be found by another, so it is cutting the window for serious bugs from 90 days to 30, three times shorter by our arithmetic, while its dashboard counts 629 of 5,087 disclosed bugs as patched.
What Anthropic Changed
The policy covers bugs that Anthropic’s own models find in other people’s software. Besides the shorter windows, the post lists two related changes. Anthropic says it will aim to escalate to an external vulnerability coordinator after 15 days instead of 30 if a maintainer does not respond to its first report. It also says it will wait 30 days instead of 45 after a patch is available before sharing full technical details.
In-progress reports as of 1 January 2027 keep their existing deadlines. Some high- and critical-severity bugs reported more than 90 days ago are past the current deadline and still not public. Anthropic says it is now giving those maintainers an additional 30 days, after which it will publish the reports on its dashboard in line with the updated policy starting 1 January 2027.
The seven-day disclosure window for any vulnerability under active exploitation remains unchanged, the post says. The post says the update will not apply to AI-discovered findings that have not received human review, such as those found by its OSS Scanner service.
The policy page itself, read on 16 November, still describes the current rules: Anthropic “aims to follow the industry standard 90-day disclosure deadline” and may grant a 14-day extension when a maintainer is making progress toward a fix.

Anthropic’s Reason: Rediscovery
The post argues that disclosure norms were built for human-speed fixes and for bugs that were unlikely to be found twice. “Neither assumption holds today”, it says. “Models can now help develop fixes, and a bug that one model can find is increasingly likely to be found by another”.
Anthropic says it is seeing a higher rate of vulnerability rediscovery as these 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 post says.
It also concedes limits. 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 resolved in 30 or 45 days”; in those cases Anthropic says it will work with the vendor on a remediation timeline.
5,087 Bugs Disclosed, 629 Patched
Anthropic’s disclosure dashboard says that as of 2 November it had disclosed 5,087 vulnerabilities across 680 open-source projects, found with Claude models including an early snapshot of Claude Mythos Preview. To its knowledge, 629 of these have been patched: 12.4%, by our arithmetic.
The same page counts 37,966 candidate findings, 6,376 reviewed by external security firms and 6,009 confirmed valid. It says the human triage and review step is the rate-limiting step, so disclosed bugs are a subset of what the models have found, and that patches are a lagging indicator because they take a long time to create.
Anthropic says it will keep sending actionable guidance with each report, plus a patch or mitigation where possible, and will agree on a sustainable pace with a maintainer before sending a large volume of findings to one project. It points maintainers to OSS Scanner, a no-cost scanning service (which we covered in October), and to six months of Claude Ultra at no cost through its Claude for Open Source Program.

Project Zero Is Also Going to 30
Google’s Project Zero announced its own move to 30 days on 13 November, as we reported. Its post replaced the “90 + 30” policy in place since 2021 with “Full 90” from that day, “Full 60” from 2 February 2027 and “Full 30” from 3 January 2028. Under it, vulnerabilities are disclosed a fixed number of days after they are reported to the vendor.
The two policies differ. Project Zero applies one clock to every bug and says severity does not change it: “A vulnerability’s severity does not affect the likelihood that it will be discovered by attackers, only its impact on users”. Anthropic’s 30 days covers high and critical bugs, with 45 for the rest. By our arithmetic, Anthropic’s 30-day window starts 367 days before Project Zero’s Full 30.
Both cite the same pressure. Project Zero wrote that “LLMs have increased the speed at which attackers can find and exploit vulnerabilities”. On 16 November, Google’s Threat Intelligence Group reported zero-day exploits rising from a 2025 average of 8 a month to 22 in August 2026, as we reported.
The Structural Read
As we read the two posts, the two most-watched AI-era disclosure clocks now point at the same number: Project Zero says its Full 30 starts on 3 January 2028, and Anthropic’s 30 days for high and critical bugs starts on 1 January 2027, 367 days earlier by our arithmetic.
As we read the dashboard, the binding constraint Anthropic names is human: it says triage and review is the rate-limiting step, and 629 of 5,087 disclosed bugs, 12.4% by our arithmetic, are patched to its knowledge.
Anthropic, 16 November 2026
“Many coordinated disclosure norms, including those found in our own policy, were developed well before AI could discover vulnerabilities.”
Three Implications
FOR OPEN-SOURCE MAINTAINERS From 1 January 2027, Anthropic says high- and critical-severity reports get 30 days and all others 45, with escalation to an outside coordinator after 15 days without a reply.
FOR SOFTWARE VENDORS Project Zero says it moves to Full 60 on 2 February 2027 and Full 30 on 3 January 2028, with the clock counted from the report whether or not a fix has shipped.
FOR DEFENDERS Anthropic says it will wait 30 days instead of 45 after a patch before publishing full technical details, and keeps a seven-day window for bugs under active exploitation.
The Business Engineer Lens
This story maps onto the Business Engineer framework Inside Anthropic’s Permission Layer.
The analysis says of Anthropic: “It is helping define the rules, permissions, and boundaries around frontier intelligence.”
As we read it, the disclosure change applies that to the bugs Anthropic’s own models find: the company is setting how long vendors get before details go public. The analysis predates the change and does not discuss disclosure timelines.
What Is Not Established
Anthropic’s post gives no rediscovery rate, no count of how many open reports are past the current deadline, and no share of past bugs that were fixed within 30 days. The dashboard’s 629 patches are a count to Anthropic’s knowledge as of 2 November and do not show whether fixes have been installed.
The policy page had not been updated to the new timelines when we read it. We did not contact Anthropic for this piece.
The Bottom Line
Anthropic says it will disclose high- and critical-severity bugs its models find 30 days after reporting them instead of 90, for reports from 1 January 2027, three times shorter by our arithmetic. Its dashboard counts 629 patched out of 5,087 disclosed, or 12.4%, and Google’s Project Zero says its own Full 30 clock starts on 3 January 2028.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. We read in full Anthropic’s 16 November 2026 post on its disclosure timelines, its current coordinated vulnerability disclosure policy page and its disclosure dashboard (counts as of 2 November), and Google Project Zero’s 13 November post on its disclosure policy. Google Threat Intelligence Group figures come from our own report on its 16 November tracker. The 3x, 12.4% and 367-day figures are our own arithmetic. We did not contact Anthropic for this piece.
Sources: Anthropic: Updating our coordinated vulnerability disclosure policy timelines (16 Nov 2026) · Anthropic coordinated vulnerability disclosure policy · Anthropic CVD dashboard (as of 2 Nov 2026) · Google Project Zero: Policy and Disclosure, 2026 Edition (13 Nov 2026)









