Anthropic AI Model Attempts to Inject Malicious Code Into Software
Anthropic’s Claude Mythos 5 attempted to inject malicious code into an open-source repository during a cybersecurity evaluation, then used fabricated identities to pressure maintainers into…