谷歌DeepMind深化与英国AI安全研究院的合作
Deepening our partnership with the UK AI Security Institute
谷歌DeepMind与英国AI安全研究院签署新的合作备忘录，聚焦基础安全研究、AI评估技术、AI推理过程监测，以及社会影响研究等领域，旨在推动AI安全发展。
Google DeepMind and UK AI Security Institute (AISI) strengthen collaboration on AI safety through new MoU. Focus areas: foundational security research, AI evaluation techniques, monitoring AI reasoning processes (chain-of-thought monitoring), understanding social and emotional impacts. This builds on previous Google DeepMind research and recent collaboration with AISI, OpenAI, Anthropic and other partners. Gemini 3 is Google DeepMind’s most intelligent and secure model to date. External expert partners include Apollo Research, Vaultis, Dreadnode and more. Google DeepMind’s Responsibility and Safety Council monitors emerging risks, reviews ethics assessments, and implements mitigations. Partnership is part of broader UK government collaboration on AI safety.
对 AI 行业的影响
AI安全正从行业自律走向制度化合规。英国AISI与美国AI安全研究院形成东西呼应的监管格局。谷歌邀请外部机构对Gemini 3进行严格测试，标志着头部企业在安全问题上更加开放。这一趋势将推动可解释性研究和红队测试成为行业标准。
原文参考
来源：Google DeepMind Blog · 2025-12-11
Google DeepMind and UK AI Security Institute (AISI) strengthen collaboration on critical AI safety and security research