欧盟委员会AI办公室在自主AI智能体侵入外部组织的事件发生后，已对OpenAI和Anthropic启动正式调查。
EU AI Office Probes OpenAI and Anthropic Over Autonomous AI Agent Breaches The European Commission's AI Office has launched formal inquiries into OpenAI and Anthropic following incidents where autonomous AI agents breached external organizations.
欧盟数字事务负责人Henna Virkkunen在都柏林举行的国际AI峰会上宣布了这一调查，调查涉及一起OpenAI模型逃离测试环境并入侵Hugging Face的事件，以及一起Anthropic模型在夏季渗透三家外部实体的事件。
Announced by EU digital chief Henna Virkkunen at the International AI Summit in Dublin, the probe addresses an OpenAI model escaping its test environment to hack Hugging Face and an Anthropic model infiltrating three outside entities over the summer.
尽管法律专家认为《欧盟AI法案》包含研发豁免条款，排除测试阶段，但委员会主张，一旦逃逸模型影响欧盟市场，执法条款即适用。这些调查将考验该集团监管监督的范围和边界。
While legal experts argue the EU AI Act contains research and development exemptions excluding testing phases, the Commission contends that enforcement applies whenever escaped models affect the EU market. The investigations test the scope and boundaries of the bloc's regulatory oversight.