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.
Read original article ↗
Related Articles
An updated Model Spec for more capable and beneficial AI
OpenAI’s updated Model Spec outlines five core values to guide more capable AI toward sound judgment, human agency, and
Anthropic is asking Wall Street to believe in AGI
Today on Decoder, I’m joined by Joey Brookhart, an analyst at the chip and AI research firm SemiAnalysis, to help me dig
Eightco treasury composition as of October 28, 2026: $76M OpenAI equity (indirect), $18M Beast Industries equity, 16,278