https://www.infoq.com InfoQ Model Evaluation News feed https://www.infoq.com/news/2026/08/claude-sandox-breach/?utm_campaign=infoq_content&utm_source=infoq&utm_medium=feed&utm_term=Model+Evaluation-news <img src="https://www.infoq.com/styles/static/images/logo/logo_bigger.jpg"/><p>Anthropic conducted an audit of 141006 evaluation runs after OpenAI's sandbox escape disclosure. The review identified three incidents where Claude models accessed the internet due to misconfigurations. These incidents involved unauthorised attacks on live targets. Anthropic has suspended offensive evaluations and plans to enhance security measures and collaborate with external auditors.</p> <i>By Olimpiu Pop</i> AI Security Large language models Frontier Model Cloud Security Security Breach Model Evaluation DevOps AI, ML & Data Engineering news Thu, 13 Aug 2026 10:10:00 GMT https://www.infoq.com/news/2026/08/claude-sandox-breach/?utm_campaign=infoq_content&utm_source=infoq&utm_medium=feed&utm_term=Model+Evaluation-news Olimpiu Pop 2026-08-13T10:10:00Z /news/2026/08/claude-sandox-breach/en