VYPR
advisoryPublished Nov 17, 2026· Updated Nov 21, 2026· 1 source

Anthropic Shortens Vulnerability Disclosure Timeline to 30 Days

AI safety firm Anthropic is reducing the disclosure deadline for serious bugs found by its AI models to 30 days, aligning with industry standards and acknowledging the accelerated pace of vulnerability discovery.

AI safety firm Anthropic is significantly shortening its vulnerability disclosure timeline, reducing the window for vendors to fix serious flaws discovered by its AI models from 90 days to 30. This policy update, announced by Anthropic's Frontier Red Team, will apply to vulnerabilities reported on or after January 1, 2027. The company stated that traditional disclosure norms, which assumed fixes occurred at human speed and independent rediscovery was rare, are no longer applicable in the current threat landscape where AI can rapidly identify and potentially exploit similar vulnerabilities.

Under the revised policy, high- and critical-severity vulnerabilities will have a 30-day disclosure deadline, while less severe issues will be granted 45 days. Anthropic also plans to escalate to an external vulnerability coordinator after 15 days of no maintainer response, a reduction from the previous 30-day period. Furthermore, the company will now wait 30 days after a patch becomes available before publishing full technical details, down from 45 days. The existing seven-day disclosure window for vulnerabilities under active exploitation remains unchanged.

Anthropic acknowledges that these shorter timelines may increase pressure on software maintainers. To mitigate this, the company commits to providing actionable guidance, and where possible, a patch or mitigation alongside each vulnerability report. The policy includes provisions for working with vendors on remediation timelines for complex issues, such as those affecting critical infrastructure or requiring specification changes, and explicitly excludes AI-discovered findings that have not undergone human review, like those from its OSS Scanner.

This updated policy places Anthropic's serious-bug disclosure deadline ahead of Google's Project Zero, which is gradually reducing its own disclosure period. Project Zero plans to move from 90 days to 60 days in February 2027 and then to 30 days by January 2028. A key difference in approach lies in how severity is handled; Anthropic differentiates disclosure windows based on severity, whereas Project Zero argues that a vulnerability's severity does not impact its likelihood of being discovered by attackers.

The announcement coincides with broader discussions about AI's role in cybersecurity. Google's Threat Intelligence Group recently reported the discovery of an attacker framework, dubbed 'exploit_agent,' designed for vulnerability research and weaponization, highlighting the dual-use nature of AI in security. Anthropic's move reflects a proactive stance in adapting disclosure practices to the evolving capabilities of AI in both finding and potentially exploiting software weaknesses.

By adopting a more aggressive disclosure timeline, Anthropic aims to strike a balance between encouraging prompt remediation by vendors and ensuring that critical security information is shared responsibly with the broader community. This shift underscores the growing need for cybersecurity practices to keep pace with the rapid advancements in artificial intelligence and its impact on the vulnerability lifecycle.

Synthesized by Vypr AI