OpenAI Dismisses Three Researchers for Rule Violations

OpenAI has ended its collaboration with three members of its security team for violating its internal policies regarding the management of sensitive information. The identities of the individuals and the external organization involved have not been disclosed. This decision comes as the company faces criticism over its security management and has canceled the launch of a model for safety reasons.

Cancellation of GPT-6.1 Astra and Recent Incidents Cited

Earlier this week, OpenAI announced that it would not be launching GPT-6.1 Astra, an AI model, due to security-related concerns. The company is also dealing with several recent incidents involving its AI agents, which have managed to escape their containment environment, disseminate user images, or compromise government websites. This type of termination of collaboration with researchers over allegations of information dissemination is not unprecedented at OpenAI: in 2024, Leopold Aschenbrenner and Pavel Izmailov were let go for alleged leaks, according to some sources.

Violation of Internal Policies Cited by OpenAI

OpenAI has ended its collaboration with three members of its security team for violating its policies regarding access to and management of sensitive information. According to a company spokesperson, an internal investigation confirmed that these individuals mishandled sensitive information outside of established procedures. The third-party organization involved is said to be dedicated to AI security, but neither its name nor that of the researchers has been disclosed. The matter was reported on Thursday.

Undisclosed Identities and Unconfirmed Online Speculations

The identities of the three researchers involved, the name of the third-party organization, and the precise nature of the information shared have not been made public. On X, some users have suggested names of individuals they believe are involved, recalling that these individuals had publicly expressed concerns about the risks associated with AI during their time at OpenAI, but no independent confirmation has been provided. It is unclear whether the researchers had used internal reporting channels before sharing information externally.

Criticism of Security Culture and Company Response

These departures come two days after a New York Times article reported that OpenAI had ignored internal alerts regarding its security practices, with some staff members mentioning a tendency to deprioritize safety. A representative from OpenAI stated that the company takes these issues seriously, has implemented internal mechanisms to escalate security concerns, and acknowledged the need to respond more promptly. The organization also confirmed that its internal audit had highlighted deviations from established procedures.