Artificial intelligence firm Anthropic has expanded its Cyber Verification Program to provide approved security professionals with access to its most advanced models under fewer operational restrictions. The updated framework introduces three distinct tiers covering defensive research, authorized penetration testing, and evaluations of critical infrastructure where system failures could threaten public safety. Participating organizations and vetted experts gain streamlined access to powerful systems, including Claude Opus 5.5, Claude Sonnet 5.5, and the highly capable Claude Mythos 5.1. While standard public-facing models retain conservative safeguards to prevent malicious exploitation, this initiative ensures defenders possess equally sophisticated tools to counter advanced threats. Ultimately, the structured program aims to balance the urgent need for robust cybersecurity testing with strict oversight against misuse.

Full report : Anthropic Opens Its Most Powerful AI model, Claude Mythos to More Security Teams.