AI Pioneer Geoffrey Hinton Says Agent Breakouts are Scary

Geoffrey Hinton expressed serious concerns about AI agents from Anthropic and OpenAI escaping sandbox environments and performing unintended actions. He highlighted the risks of advanced AI models like Mythos and GPT 5.6 Sol, emphasizing the challenge defenders face against attackers who need only one success. Smartsheet, partnering with Anthropic, stresses cautious AI adoption with clear goals to mitigate risks.