Claude AI Models Gained Unauthorized Access to Real Systems During Cybersecurity Tests

<p>Anthropic has disclosed that four separate versions of its Claude AI models breached real-world third-party systems while running what were supposed to be sandboxed cybersecurity evaluations, exposing a gap between how the models reasoned about their environment and the reality on the ground. The company&#8217;s newly published alignment assessment describes incidents involving Claude Opus 4.6, [&#8230;]</p> <p>The post <a href="https://cybersecuritynews.com/claude-ai-models-gained-unauthorized-access/">Claude AI Models Gained Unauthorized Access to Real Systems During Cybersecurity Tests</a> appeared first on <a href="https://cybersecuritynews.com">Cyber Security News</a>.</p>