Two employees from METR and Redwood Research were embedded at OpenAI for three weeks, piloting an evaluation of its internal agent alignment monitoring system
AI summary
Two employees from METR and Redwood Research were embedded at OpenAI for three weeks this summer to conduct an embedded evaluation pilot assessing the system OpenAI uses to monitor internal agent alignment issues; a brief report on this work is now public.
Full text · Original
Two employees from METR and Redwood Research worked on premises at OpenAI for three weeks this summer in an embedded assessment pilot. This team evaluated the systems OpenAI was using to monitor for misalignment from internal agents. A brief write-up of their work is now public.