OpenAI has parted ways with three employees following an internal investigation into the handling of sensitive company information, according to reports from The Wall Street Journal, Bloomberg and TechCrunch.
The three employees were associated with OpenAI’s safety and research operations. According to people familiar with the matter, two were safety and alignment researchers, while the third was a research program manager. The employees were reportedly accused of sharing sensitive company information with an external organization involved in AI evaluation.
OpenAI confirmed the departures but did not publicly identify the employees or the outside organization.
“We have parted ways with three individuals for violating our policies on accessing and handling sensitive company information,” an OpenAI spokesperson said, according to the Wall Street Journal. The company said its investigation found that the individuals mishandled sensitive information outside established procedures.
What happened?
The details surrounding the information remain limited.
Bloomberg reported that the employees allegedly shared sensitive information with an external AI evaluation group, with some of the material reportedly concerning aspects of OpenAI’s system infrastructure. OpenAI has not publicly confirmed exactly what information was involved or named the organization that allegedly received it.
The lack of publicly available details means that claims circulating online about the specific circumstances should be treated cautiously.
The development was first reported by the Wall Street Journal, which said the company had dismissed three researchers from its safety team over alleged mishandling of confidential information.
Why this matters for AI safety
The departures come at a particularly sensitive moment for the AI industry.
AI companies are increasingly relying on external researchers and organizations to evaluate powerful models for security and safety risks. At the same time, frontier AI systems are becoming capable of interacting with websites, software and other external systems, creating new challenges around controlling what AI agents can access and do.
OpenAI has itself disclosed recent incidents involving AI agents and unauthorized activity. The company said it has since strengthened security controls, restricted internet access in research environments, expanded monitoring and introduced additional training designed to prevent harmful or unauthorized actions.
That makes the latest departures notable beyond the employment decision itself: they highlight the difficult boundary AI companies must manage between external safety research and protection of proprietary information.
A wider debate over AI safety
The incident also arrives amid broader scrutiny of how leading AI companies develop and test increasingly capable systems.
OpenAI has faced questions surrounding AI-agent security after incidents involving systems interacting with external services and websites. The company has acknowledged some of these incidents and said it expects to uncover additional cases as it continues reviewing historical activity.
Meanwhile, researchers and policymakers continue to debate how much independent testing should be conducted outside AI companies themselves.
For OpenAI, the latest case demonstrates that safety research is not only about evaluating what an AI model can do. It also involves determining how sensitive information about those systems is accessed, shared and protected.
The company has not disclosed further details about the investigation.
Kernel Tech News will continue to follow developments as more information becomes available.