Opus 4.7 Cybersecurity Safeguards Block Legitimate Pentesters — GitHub Issue and Cyber Verification Program

GitHub·medium signal

A GitHub issue on the Claude Code repo documents that Opus 4.7's tightened cybersecurity filters are blocking authorized security research that worked under Opus 4.6, including drafting, analysis, and PoC refinement within approved bug bounty scopes. Anthropic responded by announcing a Cyber Verification Program for legitimate security researchers, red-teamers, and penetration testers to apply for reduced restrictions. The tension between AI safety and utility is now directly impacting the security practitioner community.