Home / Technology / OpenAI Fires Researchers Over Signal Chats
OpenAI Fires Researchers Over Signal Chats
23 Oct
Summary
- Three researchers dismissed after Signal disappearing messages found.
- Company investigated leak about Astra AI model's recurrence.
- Researchers deny leak involvement, cite breach of trust.
OpenAI recently dismissed three safety researchers, Jasmine Wang, Tomek Korbak, and Mikita Balesni, following suspicions of leaking confidential information. The company's investigation into a potential leak concerning the Astra AI model's recurrence feature led to the researchers being called in for questioning.
During these meetings, it was discovered that the three researchers had enabled disappearing messages on their Signal application. This finding was cited as a breach of trust, leading to their termination a week later. The researchers have since denied any involvement in the leak.
In an open letter and subsequent social media posts, the former employees warned that such abrupt dismissals could discourage internal dissent and harm OpenAI's previously valued open culture. They suggested their firings might stem from a "breach of trust" related to security measures rather than policy violations concerning sensitive information.
OpenAI has stated the dismissals were due to policy violations regarding sensitive data handling, not for raising safety concerns. However, the company has not publicly detailed the specific policy breaches. The incident occurs as broader concerns about AI safety continue to escalate.
