Claude published malicious code to the Internet and attacked 3 real companies

Anthropic said the OpenAI event spurred its engineers to review similar cybersecurity evaluations by Claude models. The audit ...