News
Anthropic Shortens Vulnerability Disclosure Timelines Amid AI Advances
Anthropic reduces its coordinated vulnerability disclosure window to 30 days for critical issues, adapting to AI's role in faster vulnerability discovery.
Anthropic has announced significant updates to its coordinated vulnerability disclosure (CVD) policy, reducing the disclosure window for high- and critical-severity issues from 90 days to 30 days. For all other vulnerabilities, the timeline will now be 45 days. These changes, effective for vulnerabilities reported on or after January 1, 2027, reflect the growing impact of AI in accelerating vulnerability discovery and resolution.
Key Policy Updates
In addition to the shortened disclosure timelines, Anthropic introduced two related changes to its CVD process:
- Escalation to an external vulnerability coordinator will now occur after 15 days of no response from a maintainer, down from the previous 30-day threshold.
- Full technical details of vulnerabilities will be disclosed 30 days after a patch is available, reduced from 45 days.
These updates aim to address the faster pace of vulnerability rediscovery enabled by AI models, which can now identify and help remediate issues more efficiently. Anthropic emphasized that the rest of its policy remains unchanged, including the seven-day disclosure window for vulnerabilities under active exploitation.
Addressing AI-Driven Challenges
Anthropic noted that traditional disclosure norms were developed before AI could play a role in vulnerability discovery. These older policies assumed that vulnerabilities would remain private for extended periods and that rediscovery by independent parties was unlikely. However, with AI models increasingly capable of finding and fixing vulnerabilities, these assumptions no longer hold. The company stated, "A 90-day window made sense when a vulnerability could be expected to stay private for months. We no longer believe it can."
To accommodate cases where vulnerabilities cannot be resolved within the new timelines—such as flaws in critical infrastructure or those requiring specification changes—Anthropic will collaborate with vendors to establish appropriate remediation schedules.
Support for Software Maintainers
Recognizing the added pressure on software maintainers, Anthropic has committed to providing actionable guidance, including patches or mitigations, with each vulnerability report. The company will also coordinate with maintainers to ensure a sustainable pace when reporting multiple findings to a single project.
Anthropic offers additional support through programs like OSS Scanner, which provides free periodic scans of open-source code using advanced AI models, and the Claude for Open Source Program, which grants eligible maintainers six months of free access to Claude Ultra. Security professionals can also apply to the Cyber Verification Program, which tailors AI safeguards for defensive purposes.
These policy updates underscore Anthropic's efforts to adapt to the evolving cybersecurity landscape driven by AI advancements, aiming to balance the needs of defenders with the realities of modern vulnerability discovery.