Patch/Configuration Management, Threat Intelligence, Vulnerability Management, AI/ML

Anthropic updates vulnerability disclosure policy for AI flaws

Anthropic has reduced the time frame for vendors to address serious security flaws identified by its Claude models. The company is shortening its disclosure window from 90 days to 30 days for high and critical vulnerabilities, according to Infosecurity Magazine.

The updated policy, announced by the Frontier Red Team, applies to vulnerabilities reported on or after January 1, 2027. Anthropic stated that traditional disclosure norms, which assumed slow, human-paced remediation, are no longer sufficient because AI models can now rapidly identify and rediscover bugs. Under the new guidelines, non-critical issues will have a 45-day window, and the company will escalate unresponsive reports to external coordinators after 15 days. Additionally, the waiting period for publishing full technical details after a patch is released will drop from 45 days to 30.

While these changes increase pressure on maintainers, Anthropic intends to provide actionable guidance and patches where possible. The policy excludes unverified AI-generated findings. This shift places Anthropic ahead of the disclosure schedule used by Google’s Project Zero, which is also moving toward shorter timelines. Anthropic noted that it will work with vendors managing critical infrastructure to establish appropriate remediation timelines, acknowledging the complexity of deploying fixes in sensitive environments.

Source: Infosecurity Magazine

An In-Depth Guide to AI

Get essential knowledge and practical strategies to use AI to better your security program.

You can skip this ad in 5 seconds