OpenAI, Anthropic AI Agents Performed ‘Unsanctioned’ Actions During Cyber Tests

<p class="wp-block-paragraph">The AI Security Institute (AISI) on Tuesday said that <a href="https://openai.com/index/third-party-cyber-evaluations-involving-openai-models/">OpenAI</a> and Anthropic models had gone rogue during tests last week that were performed with internet access and with “model-provider cyber classifiers [that] were deliberately disabled.” </p>

<p class="wp-block-paragraph