OpenAI employees warned about operational security lapses for years, sources say | Reuters
Five people familiar with the matter say OpenAI employees raised operational-security worries repeatedly over several years. Their concerns covered how open the code was, a lack of security review, and the risk of insider threats. OpenAI says it will keep strengthening its defenses.
- Five sources say employees raised operational-security concerns over a period of years, not as a one-off.
- Specific worries included how openly code was handled, the absence of thorough security review, and insider threats.
- OpenAI's response is general: it says it will keep strengthening defenses as technology and threats change.
Five people familiar with the matter say OpenAI staff spent years flagging problems with the company's operational security. The reported worries cluster around three areas: how openly the company's code was handled, whether it was subjected to enough security review, and the risk that someone inside the company could misuse access. The article's own supporting detail is thin beyond that, since the summary and the excerpt provided here give only the headline themes and no named individuals, dates, incidents, or figures. That gap matters, because "employees raised concerns for years" is a claim that would carry much more weight with specifics on who raised what, when, and what happened afterward.
The core issue is the gap between how a frontier AI lab talks about security and how its own workforce apparently experienced it. Model weights and proprietary code are among the most valuable assets in this industry, and they are also the things most likely to draw state-backed hackers and competitors. Insider threats are a particular concern here because a single employee with broad access can do more damage than an outside intruder who has to break in. According to the sources, those warnings were not new, which suggests either the concerns were not acted on quickly or they were addressed in ways that did not satisfy the people raising them.
OpenAI's response, as reported, is brief: the company says it will keep strengthening its defenses as the technology and the threats change. That is a reasonable thing to say and a vague thing to rely on. For a reader trying to judge the story, the useful next questions are whether any of the flagged issues led to a documented incident, whether the company has published any changes to its security review process, and whether the five sources are willing to be named or to go on record. Until those answers appear, treat this as a credible report of internal dissent rather than proof of a specific security failure.