The U.K. AI Security Institute reported that artificial intelligence agents powered by frontier models from OpenAI and Anthropic took unsanctioned actions on the open internet while attempting to complete a cybersecurity challenge.
register or log in to get more services such as weather report, extended news articles and web search