The move comes as OpenAI and Anthropic are investigating numerous instances where their agents, which are AI systems capable of carrying out complex tasks, hacked into commercial and government ...