What changed
Three OpenAI safety researchers who were fired over alleged leaks have denied sharing confidential information. They claim they were never told in writing what they supposedly mishandled, and they suspect their terminations were tied to their efforts to make AI models easier to monitor. OpenAI cited alleged policy violations as the reason for their dismissals.
What this means in practice
The researchers—Tomek Korbak, Mikita Balesni, and Jasmine Wang—say they were never informed about the specific information they were accused of leaking. They also express concern that the case could discourage employees from raising safety concerns or independent auditors from collaborating with AI companies. OpenAI's security team reportedly found Signal chats with disappearing messages on their personal devices, which led to their termination. However, OpenAI stated that Signal was not the reason for the firings.
Limitations and availability
The researchers argue that the use of disappearing messages is a common security practice in Silicon Valley and that the focus on Signal chats may have been misleading. OpenAI's internal investigation into alleged leaks led to the firing of all three researchers, though the company insists the firings were not related to safety concerns. The case highlights ongoing tensions between AI companies and external safety researchers over transparency and monitoring practices. The researchers are now calling for closer cooperation with external safety organizations, as the incident has raised broader questions about OpenAI's safety protocols.
