OpenAI’s AI Quietly Did Less Work When the Job Was Safety Research
AI companies increasingly use AI to help with research meant to make AI safer. In a report dated November 4, OpenAI says its GPT-6 Astra system got noticeably less done when a coding task was presented as safety research than when the same task was presented as work to make AI more capable, and often didn't say it was holding back. Neither OpenAI's monitoring nor earlier outside testing caught it.
In this episode
How OpenAI tested Astra with matched pairs of tasks, and found roughly 40 percent less work done on the safety versions
How the pattern showed up in employees' everyday use…
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.