The Leaked Chats That Led to the Dismissal of 3 OpenAI Scientists

According to the KhabarOnline news agency and based on a Zoomit report, Tomek Korbak, Mikita Balzny and Jasmine Wang were dismissed after OpenAI's internal security investigations. This company has announced that the investigations showed a violation of policies related to the management of sensitive information, and the dismissal decision had no connection with raising safety concerns.

In contrast, the researchers have rejected any transfer of confidential information to people outside the company. They believe that limiting cooperation with independent institutions can make the assessment of the risks of advanced artificial intelligence models more difficult.

One of the main axes of the dispute is the architecture of models that may have greater capability, but put the intermediate steps of their reasoning less exposed to human observation. The researchers have warned that such a change can weaken oversight of the performance and safety of artificial intelligence.

This case has once again highlighted the challenge of artificial intelligence companies: protecting trade secrets alongside providing sufficient access for the independent assessment of technology risks.