Anthropic Safeguards Blog

Articles

November 2026

Strengthening our safeguards with frontier automated red-teaming

We built Anvil, an automated red-teaming system that helps us strengthen our safeguards before we deploy them.
October 2026

Scaling Misuse Detection with Self-Assessment

Self-assessment lets Claude assess the safety of its own responses. It scales well with model capabilities and helps us identify emergent harms.

About the Anthropic Safeguards Blog

We are Anthropic's Safeguards team. We work to keep AI systems like Claude both helpful and safe. Our work spans safety systems, monitoring, robustness, cybersecurity, and privacy. We'll use this blog to share research notes and findings from our work. We hope they're useful to others working on similar problems.