OpenAI confirmed that it parted ways with three researchers over alleged mishandling of sensitive information, an episode that has drawn attention to internal controls at leading artificial-intelligence companies.
A company spokesperson told CBS News that the employees had handled sensitive information outside OpenAI’s procedures. OpenAI did not publicly identify the information, explain how it was handled, or provide a detailed account of each person’s conduct.
What has been reported
The Wall Street Journal first reported that the three safety researchers were dismissed for alleged misconduct, including claims that confidential information was shared with a third-party AI-safety organization. Those details remain allegations reported by the Journal; the company has not released a public evidentiary record.
That distinction matters. The confirmed fact is that OpenAI says the three employees left over information-handling issues. The nature, sensitivity and consequences of the alleged disclosures have not been fully established in public.
A separate safety-policy departure
Reuters reported October 3 that another OpenAI safety employee, David Robinson, resigned separately and criticized what he described as the company’s approach to safety and deployment. His departure is not reported as part of the three information-handling cases and should not be treated as evidence about them.
Taken together, the events show an active internal debate about how AI companies balance confidentiality, external safety scrutiny and rapid product development. They do not by themselves establish that confidential information reached a competitor or caused a security breach.
Why internal controls matter
AI laboratories hold model research, security evaluations, customer information and deployment plans that can be commercially and socially consequential. Clear access rules, whistleblower channels and procedures for working with outside organizations are therefore important for both corporate security and legitimate safety concerns.
The news arrives as regulators examine how AI companies manage consumer and public risks. AskNovus recently covered the FTC inquiry involving OpenAI, Anthropic and other AI labs. For a separate product-development story, see the restricted rollout of Google’s Gemini 4 Argon.
Until OpenAI or the former employees release more detail, claims about what was shared, why it was shared and what harm resulted should be treated cautiously rather than inferred from anonymous or secondary accounts.
Sources
- CBS News report and OpenAI statement, updated October 1, 2026.
- The Wall Street Journal report, October 1, 2026.
- Reuters report on the separate resignation, October 3, 2026.
Featured-image credit: Hitesh Choudhary/Unsplash. The image is illustrative.