"neither assumption holds today." — Anthropic
Anthropic's new disclosure timelines
Anthropic's Frontier Red Team announced on November 16 that the company is shortening the time it gives vendors and maintainers to fix serious flaws discovered through its Claude models. Beginning with vulnerabilities reported on or after January 1, 2027, Anthropic will require fixes for high- and critical-severity vulnerabilities within 30 days; all other reported issues will have a 45-day disclosure deadline. Reports already in progress on that date will retain their existing deadlines.
The company framed the change as a response to two shifting assumptions: that fixes advance at "human speed" and that independent rediscovery of flaws was unlikely. Anthropic said those assumptions no longer hold because "bugs one model could find were increasingly likely to be found by another."
Escalation, exceptions, and technical disclosure windows
The updated policy tightens several procedural elements beyond the headline deadlines. Anthropic will aim to escalate reports to an external vulnerability coordinator after 15 days without a maintainer response, rather than the prior 30-day escalation window. Where a patch is published, Anthropic will wait 30 days before releasing full technical details — down from 45 days — though the existing seven-day disclosure window for vulnerabilities under active exploitation remains unchanged.
Anthropic said it would work with vendors on remediation timelines for flaws in software that is hard to deploy — explicitly citing critical infrastructure — or issues that require specification or RFC changes. The company also committed to sending actionable guidance with each report and, where possible, a patch or mitigation to ease the shorter timelines. The policy explicitly does not cover AI-discovered findings that have not yet received human review, naming its OSS Scanner as an example.

Your scanner finds 4,000 vulns. Which 12 matter?
Nubivance is a Rapid7 Registered Partner delivering vulnerability management as a service - scanning, risk-based prioritization, and remediation follow-through across IT and OT.
Fix the backlogHow Anthropic's schedule lines up with Project Zero and Google's reporting
The company’s move puts its serious-bug deadline ahead of the schedule used by Project Zero. Google's security team currently publishes details 90 days after reporting a bug whether or not it is fixed; Project Zero will reduce that to 60 days on February 2, 2027 and to 30 days on January 3, 2028. Anthropic’s approach also differs philosophically: it sets disclosure windows by severity, while Project Zero has argued that severity does not change how likely attackers are to find a vulnerability.
On the same day Anthropic announced the change, Google's Threat Intelligence Group reported finding an exposed server hosting an attacker framework called exploit_agent, described as built to research vulnerabilities and weaponize them — a contemporaneous example, in Anthropic's framing, of accelerating rediscovery and weaponization risks.
What this means for maintainers, vendors, and external researchers
- Maintainers and vendors: They face shorter remediation timelines for high- and critical-severity flaws — 30 days — and faster escalation (15 days) if there is no response. Anthropic said it will deliver actionable guidance and, where possible, patches or mitigations to reduce friction, and it will negotiate extended timelines when a fix requires changes to specifications or is hard to deploy in critical infrastructure.
- External researchers: The policy applies to vulnerabilities Anthropic reports to vendors and maintainers, not generally to bugs reported to Anthropic by outside researchers. Anthropic also excluded AI-discovered findings that lack human review (for example, results from its OSS Scanner) from this deadline regime.
- Operators of critical infrastructure: Anthropic signaled it will work with maintainers on remediation timelines for hard-to-deploy software and for flaws needing specification or RFC changes, implying bespoke handling rather than automatic adherence to the 30- and 45-day windows.
Conclusion
Anthropic’s deadline compression — 30 days for high- and critical-severity bugs and 45 days for others, new escalation after 15 days, and a shorter post-patch silence before full disclosure — is a deliberate acceleration meant to reflect models’ faster pace of discovery and greater risk of independent rediscovery. The company acknowledged that the shorter timelines could increase pressure on maintainers and said it would supply guidance and fixes where possible, and negotiate when fixes are complex. With Project Zero moving toward comparable deadlines on a different timetable and threat intelligence groups flagging tools such as exploit_agent, Anthropic’s change crystallizes a question the company itself posed: do traditional disclosure norms match the operational tempo of today’s automated discovery and weaponization?
https://www.infosecurity-magazine.com/news/anthropic-disclosure-deadline-30/




