Anthropic Shrinks Bug Disclosure Windows to 30 Days

Anthropic is cutting the time software vendors get to fix security flaws it finds. The company has updated its coordinated vulnerability disclosure (CVD) policy, and the standard 90-day window drops to 30 days for high- and critical-severity bugs. According to Anthropic, the new timelines apply to vulnerabilities reported to vendors and maintainers on or after January 1, 2027.

In short, AI is finding bugs faster than the old system was built to handle, so the deadlines are getting tighter.

The Quick Version

  • High and critical severity: The disclosure window falls from 90 days to 30.
  • All other issues: The window falls from 90 days to 45.
  • No response from a maintainer: Anthropic will aim to bring in an outside vulnerability coordinator after 15 days instead of 30.
  • Actively exploited bugs: The 7-day window stays the same.
  • Start date: Reports filed on or after January 1, 2027.
  • Not covered: AI-discovered findings that no human has reviewed, such as output from Anthropic’s OSS Scanner.

Anthropic says the rest of the policy stays the same.

Why Anthropic Is Doing This

Coordinated disclosure is how the security world handles this problem. A researcher finds a bug, tells the vendor privately, and gives them a set amount of time to ship a fix before the details go public. For years, 90 days has been the default, largely because Google’s Project Zero made it the standard.

Anthropic’s argument is simple. Those norms, it says, were developed well before AI could discover vulnerabilities. Back then, finding a serious bug took a skilled human days or weeks of work. Attackers faced the same limits, and that gave defenders some breathing room.

That room is shrinking. If Claude can find a flaw quickly, a well-resourced attacker using similar tools probably can too. A 90-day window now means a known bug may sit unpatched for three months while others find it on their own. Shorter deadlines are Anthropic’s way of matching disclosure to how fast discovery happens now.

The Tension for Maintainers

Anthropic doesn’t pretend this is free. The company acknowledges that shorter timelines may put more pressure on software maintainers.

That matters, especially in open source. Many critical projects run on a handful of volunteers. A 30-day clock for a critical bug, on top of a growing pile of AI-generated reports, is a real burden for a two-person team working nights and weekends.

Two parts of the policy help with this:

  • Human review stays in. Unreviewed AI findings don’t trigger the new deadlines. That protects maintainers from piles of machine-generated reports that might be noise.
  • Faster escalation. If a maintainer goes silent, a coordinator steps in after 15 days. That gets more people working the problem sooner rather than just letting the clock run out.

What It Means for You

If you ship or maintain software, a few practical points:

  • Check your intake process. Make sure security reports reach a real person fast. With a 15-day escalation trigger, an unmonitored inbox now costs you more.
  • Plan for 30-day patch cycles. Critical fixes need a release path that fits inside a month, not a quarter.
  • Expect more reports. AI-assisted bug hunting is ramping up across the industry, not just at Anthropic.
  • Mark January 1, 2027. Reports filed before then keep the old deadlines.

A heads-up on one detail: some secondary coverage says the waiting period between a patch and public write-up also changed. We couldn’t confirm that in Anthropic’s own post, so check the policy page directly before you plan around it.

What Comes Next

What stands out here is the signal, not just the numbers. An AI lab is saying out loud that its own models have changed the economics of finding vulnerabilities, and it’s rewriting its security commitments to match. Other AI labs and big security teams will likely face pressure to follow, and the industry’s 90-day default could start to look outdated.

The tricky part will be capacity. Faster discovery only improves security if maintainers can actually ship the fixes. Expect more debate over funding and tooling for open source maintainers as these timelines take hold. You can find the full policy details in Anthropic’s original post.

Scroll to Top