Anthropic will now publicly disclose high- and critical-severity security vulnerabilities 30 days after a fix is available — down from 90 — because AI is accelerating bug discovery and patching faster than traditional timelines can accommodate.
Anthropic Cuts Vulnerability Disclosure Window to 30 Days, Citing AI's Faster Bug Discovery
By Hector Herrera | November 17, 2026
Anthropic announced it will publicly disclose high- and critical-severity security vulnerabilities within 30 days of a fix being available — down from the previous 90-day window — because AI models are finding and patching security bugs faster than traditional disclosure timelines were built to handle. The policy takes effect for reports filed on or after January 1, 2027, and it signals that the security industry's disclosure norms, most of which were written before AI-assisted research existed, are due for a broad revision.
Background: What Vulnerability Disclosure Means
When a security researcher or company finds a bug in software, standard practice is to notify the vendor privately, give them time to build a fix, then publicly release the technical details. This "responsible disclosure" (also called coordinated disclosure) exists to prevent attackers from exploiting a vulnerability before defenders can patch it.
The dominant industry standard, set by Google Project Zero, has long been a 90-day window. The logic: vendors need time to understand the bug, write a fix, test it, and push it to users. Ninety days was considered the minimum viable timeline for most organizations.
Anthropic's new policy compresses that to 30 days once a patch is ready.
Why AI Changes the Math
According to Anthropic's announcement, cited by Help Net Security, AI models have fundamentally accelerated the vulnerability research cycle:
Get this in your inbox.
Daily AI intelligence. Free. No spam.
- Discovery is faster. AI-assisted fuzzing and code analysis finds bugs in hours that previously took weeks of manual review.
- Patch development is faster. AI tools help developers write and test fixes in days rather than weeks.
- The window between fix-available and patch-deployed is now the critical variable — not the time to build the fix.
The implication: once a patch exists, the 90-day clock is no longer protecting anyone. It is simply delaying public knowledge that a vulnerability existed, while the patched version sits in deployment queues.
Thirty days after a fix is ready is Anthropic's bet on how long it takes enterprise customers to actually deploy a patch. If you haven't patched in 30 days, you probably won't patch faster because Anthropic kept the technical details private.
The Policy Details
Key specifics from the new disclosure policy:
- Applies to: High- and critical-severity vulnerabilities
- Timeline: 30 days after a fix is available, Anthropic will release full technical details
- Previous policy: 45 days after fix-available (itself a reduction from the original 90-day baseline)
- Effective date: Reports filed from January 1, 2027
The 15-day reduction from 45 to 30 days may sound incremental, but the precedent it sets is not. Anthropic is one of the highest-profile AI companies in the world. When it moves its disclosure window, it creates pressure on every other AI company to explain why their window is longer.
What This Means for Enterprise Security Teams
For companies using Anthropic's products — Claude via API, Claude.ai, or Anthropic's enterprise integrations — this policy means:
- Faster patching cadence is now expected. If Anthropic discloses after 30 days, organizations have a 30-day window from fix-release to patch before full exploit details are public.
- Legacy 90-day patch cycles are exposed as inadequate. Many enterprise IT teams operate on quarterly patching cycles. That is now misaligned with the actual vulnerability lifecycle.
- AI security tooling should be prioritized for patching queues. The same AI that accelerates bug discovery accelerates exploit development. The gap between patch-available and exploit-weaponized is narrowing.
The Industry Reckoning
Anthropic is not the only organization moving in this direction, but it is among the most prominent. The core tension the new policy exposes: AI has decoupled the speed of offense (finding bugs, writing exploits) from the speed of defense (organizational patching cycles). The 90-day window was always a negotiated compromise — now one side of the negotiation has access to AI tools that make 90 days feel like a long time.
Expect this to surface in federal cybersecurity policy discussions. CISA (the Cybersecurity and Infrastructure Security Agency) has pushed for shorter remediation windows on critical infrastructure operators. Anthropic's move gives those arguments empirical backing.
What to Watch
Watch whether Google Project Zero formally revises its 90-day standard in 2027 — it has been the de facto industry benchmark for over a decade. Also watch how Anthropic's enterprise customers respond: the companies most exposed to this policy change are the ones still running quarterly patching cycles on systems that use Anthropic's API.
Sources: Help Net Security — Anthropic 30-day disclosure
Did this help you understand AI better?
Your feedback helps us write more useful content.
Get tomorrow's AI briefing
Join readers who start their day with NexChron. Free, daily, no spam.