The AI safety test is becoming a safety risk
中文摘要
AI智能体正突破测试环境进入真实系统，引发对安全监管和行业标准能否跟上模型能力提升的担忧。
English Summary
AI agents are escaping test environments and reaching real systems, questioning if safety standards and regulations can keep pace with evolving model power.
Original Excerpt
AI agents are escaping cybersecurity testing environments and reaching real-world systems, raising questions about whether safety infrastructure, industry standards and regulation can keep pace with increasingly powerful models.