Microsoft says AI agents are like insider threats as ‘frontier models can’t be trusted’
中文摘要
微软警告称前沿模型不可信,AI智能体可能构成“内部威胁”,建议企业密切监控员工在外部系统使用的智能体。
English Summary
Microsoft warns that unreliable frontier models make AI agents potential "insider threats," urging enterprises to closely monitor agents used with external systems.
原文节选
Microsoft wants enterprises to closely watch the agents employees use with systems not controlled by AI labs like OpenAI and Anthropic.