SITUATION EXPLAINED: How does OpenAI monitor a model at every single token? • A new multi-stage monitoring setup that runs at every sampled token • Concerns escalate to increasingly sophisticated, high-compute automated investigators • The aim is an alert within 30 minutes, and if teams can't confirm it's a false positive in that window, they're expected to pause the activity • They paused inference in research clusters for any run that could execute code or reach the internet, then reviewed…