“Anthropic released Claude Opus 4 and subsequent models under its AI Safety Level 3 protections due to concerns about bioweapon creation capabilities.”