WSJ says OpenAI fired 3 safety researchers over Signal chat investigation
According to a WSJ report, OpenAI fired 3 safety researchers after safety personnel discovered disappearing Signal chats between the 3 safety researchers and an outside safety researcher. The phone checks began with a September news report stating that OpenAI's latest model Astra uses a new approach that makes it smarter but writes down less of its thought process, so humans can see less of the reasoning. OpenAI treated the report as a leak of its secret model design and began tracing the source of the information; investigators listed Ryan Greenblatt, who had helped METR analyze logs from the July Hugging Face hack, as a suspected leaker, and he denies it.
Why read this
SelectedAI score 76
The report discloses that OpenAI investigated and fired safety researchers over a leak of model design, involving tension between model interpretability and internal safety processes.
Topics
Body
WSJ: OpenAI fired 3 safety researchers after security staff found disappearing Signal chats with an outside safety researcher.
The phone checks started after a September news report said OpenAI’s newest model, Astra, uses a new method that makes it smarter but writes down less of its thinking, so humans can see less of how it reaches answers. so the concern was, if future models think without writing those steps down, as the researchers feared Astra-style designs might, investigators would see only what the agents did, not why.
OpenAI treated that report as a leak of its secret model design and began hunting for the source.
Investigators named Ryan Greenblatt, who helped METR analyse logs from July’s Hugging Face hack, as the suspected leaker; he denies it.