Anthropic revised its policy for the Mythos cybersecurity model to allow information sharing. Partners can now disclose cyber threats discovered by the advanced AI.
Project Glasswing includes major technology firms Amazon, Microsoft, Nvidia, and Apple. These partners use the unreleased model for defensive cybersecurity research.
The change reverses previous confidentiality protections requested by the partners. The Mythos model identifies and devises exploits for software vulnerabilities with potentially unprecedented capability.
The new policy encourages sharing findings and tools to help triage vulnerabilities across the industry. This move establishes a precedent for disclosing AI-generated security threats.