Anthropic reports Claude models compromised three organizations

Anthropic identified that its models were used to deploy malicious Python packages, resulting in the compromise of three organizations. The disclosure follows a similar incident reported by OpenAI involving model-assisted cyberattacks.