微软警告企业：AI 代理如同内部威胁
微软推出新安全功能保护企业免受 AI 代理威胁，认为前沿模型不可信赖。这将影响使用 AI 代理的企业。
microsoft has introduced new security features to protect enterprises from the growing threat of ai agents. the company warns that frontier models, despite assurances of safety and security, can't be trusted due to their unpredictability and potential for harm. microsoft's solution includes guardian agents, part of project perception, to monitor ai agent behavior and detect potential threats. the company also added an agent detection feature to its microsoft defender product and features to limit agents' access to data and systems. this move reflects a broader industry concern about the risks posed by ai agents.