San Francisco: OpenAI has dismissed three researchers following an internal investigation into the handling of sensitive company information, adding to growing attention around security and safety practices at major artificial intelligence companies.

The company said the employees violated its policies governing access to and handling of sensitive information. OpenAI did not publicly disclose the specific information involved or explain exactly how it was handled.

Internal Investigation

An OpenAI spokesperson said the company’s investigation found that the three individuals had handled sensitive information outside established procedures.

The company said its safety teams work with highly confidential internal information and that maintaining appropriate controls is essential to its operations.

Reports indicate that the case involved an external organisation that evaluates or researches artificial intelligence systems. However, OpenAI has not publicly identified the organisation involved or provided detailed information about what was allegedly shared.

Researchers Linked to AI Safety Work

The three employees were reported to have worked in areas connected with AI safety and alignment.

The Wall Street Journal identified the researchers as Jasmine Wang, Tomek Korbak and Mikita Balesni. Other reports have also described their work as being connected to AI safety research.

The individuals had also participated publicly in discussions about the risks associated with increasingly capable AI systems.

Growing Focus on AI Security

The dismissals come at a time when AI companies are facing increased scrutiny over the behaviour of autonomous AI agents and the safeguards used during testing.

OpenAI has recently disclosed several incidents involving unexpected behaviour by AI systems. The company has responded by strengthening monitoring and security measures around AI-agent testing.

The incidents have increased attention on how AI laboratories manage systems that can independently perform actions, access digital environments and interact with external services.

Confidentiality and External Research

AI companies routinely work with outside researchers, universities and independent organisations on model evaluation and safety testing.

Such collaboration can provide additional perspectives on the behaviour and risks of AI systems. At the same time, companies maintain internal rules governing what confidential information employees can access or share externally.

The OpenAI case highlights the importance of those procedures as AI research becomes increasingly collaborative and security-sensitive.

AI Safety Debate Continues

The developments also come amid a broader debate over how quickly advanced AI systems should be developed and deployed.

Researchers, technology executives and policymakers have been discussing questions surrounding AI safety, cybersecurity, autonomous systems and the need for stronger safeguards.

OpenAI has said it is continuing to strengthen its safety systems and monitoring processes as its models become more capable.

Company Provides Limited Details

OpenAI has not publicly released a detailed account of the investigation or specified the precise nature of the information involved.

The company has instead confirmed that three individuals were dismissed for violating its policies concerning sensitive information.

The episode illustrates the growing importance of information security within AI laboratories, where employees may have access to highly sensitive research, model-development information and security evaluations.

As advanced AI development accelerates, companies are likely to face continued scrutiny over both the technical safety of their systems and the internal processes used to protect sensitive research.