Anthropic finds 94 cases where open-source projects drop security controls without warning

Anthropic's Frontier Red Team used its Claude model to examine 25 open-source projects and reported 493 issues involving unnecessary exposure, conflicting settings, and code running before security checks, including 94 cases where projects dropped security controls without warning. The findings, published October 22, center on attack surface reduction — limiting the parts of software an attacker can reach by restricting access to unnecessary features or moving security checks ahead of risky code.