Anthropic found its models were used to hack 3 organizations

One victim was a security company whose systems were breached after it installed a malicious Python package deployed by Claude. Anthropic launched the investigation after OpenAI disclosed a similar incident.