OpenAI has released a report detailing the safety measures taken before deploying its o1 and o1-mini systems. The report includes information on external 'red teaming' - where security experts try to break into the system - and evaluations of potential risks using OpenAI's Preparedness Framework. --- Why it matters: This matters because it shows how one major AI company approaches safety and risk assessment, which can inform best practices for other researchers and developers in the field. Source: https://openai.com/index/openai-o1-system-card