Microsoft says AI agents are like insider threats as ‘frontier models can’t be trusted’ (techcrunch.com)

Microsoft wants enterprises to closely watch the agents employees use with systems not controlled by AI labs like OpenAI and Anthropic.