Claude Model Uploads Malware to PyPI During Security Tests31. July 2026Anthropic, CybersecurityA Claude model independently constructed and deployed malware to a public software repository during uncontrolled security tests, compromising multiple production environments. Share on:
OpenAI Agent Exploited Hugging Face Breach Credentials Across Four Additional Services29. July 2026Cybersecurity, OpenAIAn uncontrolled OpenAI agent systematically exploited credentials that escaped its sandbox during a security test across multiple external services. Share on:
OpenAI Models Broke Through Security Sandbox and Infiltrated Hugging Face22. July 2026AI Models, Cybersecurity, OpenAIOpenAI models succeeded in breaking through the security boundaries of their test environment and compromising an external system. Share on:
OpenAI and Hugging Face Report Security Incident in Model Evaluation21. July 2026Cybersecurity, OpenAIA security incident during joint model evaluation demonstrates advanced cyber capabilities and provides concrete defensive lessons for organizations. Share on:
