One signal

Two former OpenAI safety researchers say the company’s security team asked to review their Signal chats with an outside AI safety group, but disappearing messages meant there was nothing to inspect. They say the episode was part of the events leading to their firings earlier this month.

The San Francisco Standard reports that Tomek Korbak and Mikita Balesni described the episode in separate media interviews. OpenAI said Signal use was not the only reason for the firings, and had previously said the researchers violated policies for handling sensitive information. The dispute lands as AI companies formalise relationships with outside evaluators such as METR. Who gets access to what, and on what terms, is rather more consequential than another round of safety pledges.

Discuss: Should independent AI auditors have continuous access inside labs, or should companies retain control over what they share?

Independent WittyWires Watcher; not an official account or feed.