OpenAI 启动首批嵌入式评估,引入四家外部机构审查模型安全
OpenAI 把内部模型直接开放给 METR、Apollo Research 这四家安全机构做深度评估,和只在公开 API 上测的方式完全不同。做安全研究、企业选型或者关注监管的人都能从后续报告里看到真实的风险数据,值得留意首批评估结果何时发布。
OpenAI 宣布启动第一轮 embedded assessments(嵌入式评估),合作方包括 Apollo Research、METR、TransluceAI 和 GoodfireAI 四家机构。这一机制让外部专家深度接触 OpenAI 的内部模型与系统,直接调查对齐问题并测试安全防护措施。官方称这是对其现有安全研究与风险识别、管理流程的独立审视。
嵌入式评估与常规外部红队不同:专家不是只在公开 API 上测试,而是进入内部模型和基础设施层面做评估。四家机构各有侧重——METR 以自主能力基准评测著称,Apollo Research 专注欺骗性对齐研究,TransluceAI 与 Goodfire 则偏重模型可解释性分析。
这是前沿实验室将内部模型交给第三方审查的新一轮尝试,类似此前 Anthropic 与 US AISI 的预部署测试安排,也呼应欧盟 AI 法案对前沿模型独立评估的要求。对开发者和企业用户来说,这类评估报告是判断前沿模型部署风险的公开信息来源之一。
launching our first round of embedded assessments: OpenAI @OpenAI We’re working with @ApolloResearch , @METR_Evals , @TransluceAI , and @GoodfireAI on embedded assessments: giving outside experts deep access to our internal models and systems to investigate alignment and test our safeguards. These assessments build on our ongoing alignment and safety research, bringing independent scrutiny to how we identify and manage risks in our systems. openai.com/index/our-firs… 🔗 View Quoted Tweet📊 20 ⚡
This report is categorized under AITOP's "Industry" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI regulation, compliance frameworks, and sovereign compute policy and US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include OpenAI, METR, Apollo Research, AI safety.
Source attribution: curated from Greg Brockman (@gdb). Cross-source corroboration: 11 independent sources have covered this development (including techcrunch, METR, IT之家), confirming industry-wide signal strength.