OpenAI
How we detect frontier model misalignment
OpenAI uses layered monitoring to detect frontier model misalignment and unauthorized agent behavior so responders can intervene before incidents escalate.
Summary from the publisher's feed.
More in AI labs
- 12 NovHow Amplitude ships product improvements in 1–2 days with OpenAIOpenAI
- 12 NovGoogle Search gets a major upgrade with Gemini 4 HydrogenGoogle AI
- 12 NovHoliday Extras is rethinking what it can build with CodexOpenAI
- 12 NovSupporting research on AI and children’s safetyOpenAI
- 11 NovGemini Decisions API: Fast, calibrated multimodal decisionsGoogle AI
- 11 NovTry Tabletop: Customizable, collaborative spaces for groups to plan, play and do more together.Google AI
- 11 NovAlphaGenome 2: Deciphering how cells read DNA to better understand diseaseGoogle DeepMind
- 11 NovGilbert + Tobin rethinks the business of law with AIOpenAI