OpenAI dismissed three researchers on October 1, after an investigation concluded they mishandled sensitive information outside established procedures. The firings at the San Francisco-based artificial intelligence lab occurred amid an intense industry debate over model safety and existential risks.
OpenAI Dismisses Three Researchers Over Sensitive Information
OpenAI confirmed the dismissals on October 1, stating that an internal inquiry found the employees violated company policies. The ChatGPT creator did not confirm their identities in its initial statement, though external reporting identified at least two of them as working on safety and alignment.
Company representatives stated that an internal review determined these individuals handled confidential data outside approved company protocols, breaking foundational trust. According to statements provided to media outlets, the company parted ways with three people for breaching rules regarding the access and management of restricted corporate details.
While public speculation connected the departures to recent public criticisms voiced by tech employees, the BBC understands the former employees were not let go for raising safety concerns, but specifically for the alleged mishandling of sensitive data involving work with an external organization that evaluates AI models.
Researchers Warn That Rushing AI Development Gambles with Lives
The departures coincide with an active public discourse among artificial intelligence researchers regarding recursive self-improvement and safety guardrails. Several researchers recently posted public warnings online concerning the pace of advanced model development. In September, a 27-year-old researcher named Jacob Coxon resigned from Anthropic, cautioning that major AI laboratories were gambling with human lives by rushing toward the creation of increasingly powerful systems.
Identified former employees Jasmine Wang, Tomek Korbak, and Mikita Balesni had regularly posted about AI safety-related issues on X in the weeks leading up to the disciplinary action. Mikita Balesni shared a post on September 10 expressing a personal estimate regarding existential threats, while Tomek Korbak commented publicly about being deeply unhappy with several aspects of OpenAI’s operations.
Amid these staffing changes and internal shifts, product developments at OpenAI have also adjusted. The organization cancelled the rollout of a new model known as Astra 6.1 after finding it unreliable and prone to ignoring user prompts. Instead, the firm debuted an updated version of a different model, titled GPT-6.1 Sol, during its annual DevDay conference held on September 29 in San Francisco.
Trump Signs Morally Binding Agreement with Technology Executives
The personnel changes arrived during a busy week for artificial intelligence policy in the United States. President Donald Trump hosted a meeting at the White House with top technology executives, including leaders from OpenAI, Anthropic, Nvidia, SpaceX, Meta, and xAI, to address artificial intelligence topics.
Following the gathering, a document was released that was described as a morally binding
agreement intended to act as a form of protection against potential technological dangers. Critics of the pact noted that it leaves artificial intelligence firms to regulate themselves, while President Trump has frequently downplayed alarms regarding the risks of the technology.
Concurrently, broader inquiries into the industry continue to develop. The Washington Post reported on September 30 that the Federal Trade Commission initiated a broad investigation into artificial intelligence safety practices at both OpenAI and Anthropic, though the precise scope of that inquiry remains unspecified.