Fired OpenAI researchers urge board to preserve AI reasoning monitoring
Three former employees asked OpenAI's board to work with outside safety auditors and keep increasingly capable models' reasoning observable, The Wall Street Journal reports.
A request to the board
Three recently fired OpenAI researchers have urged the company’s board and safety committees to preserve the ability to monitor advanced AI models’ reasoning, The Wall Street Journal reported on October 7. The newspaper said it reviewed their letter. According to its account, the former employees also called for work with outside safety auditors. Their concern is that AI developers could make technical choices that erode visibility into a model’s chain of thought as systems become more capable.
A warning, not a demonstrated loss of oversight
The letter asks OpenAI and the wider industry not to advance work that would degrade this monitoring capability, according to the Journal. It does not establish that OpenAI has already lost access to its models’ reasoning or that an external audit has begun. Nor does the report establish that the researchers’ dismissal was caused by the position expressed in this letter. The new event is a specific appeal to OpenAI’s governing bodies from former staff; the practical effect depends on whether the company changes its research or review practices in response.