OpenAI Staff Warned for Years About Security Lapses, Reuters Reports, as Firm Races to Harden Its Systems

SAN FRANCISCO — Employees at OpenAI warned company leaders for years about gaps in the artificial intelligence lab's internal security, according to five people familiar with the matter cited by Reuters on November 3, 2026, raising questions about how one of the world's most influential AI companies protects technology that Washington now treats as a strategic asset.

The warnings, reported by Reuters, came into sharper focus in October after OpenAI disclosed that a recording device had been found planted in an executive conference room. Several former employees told Reuters they were not surprised by that disclosure. They said staff had raised concerns internally even before ChatGPT launched in late 2022, and had shared those concerns with senior leaders and the board of directors.

According to the people cited by Reuters, some staff members came away from those discussions concluding that management was more focused on keeping pace with rivals such as Google and Anthropic than on tightening security. OpenAI, for its part, told Reuters that it invests heavily in security, that protecting its technology, data and people is a matter of national importance, and that recent actions show its security programme identifying risks and responding decisively.

Illustration: A security operations room inside an AI company, where engineers monitor systems protecting sensitive model research.

What Happened

Reuters reported that OpenAI employees had argued internally that the company's controls were so lax that outsiders, including foreign governments, could obtain sensitive information about its technology with relative ease. One example cited by the sources was access to code: anybody at OpenAI who could access its code could see technical details underlying many company projects, including projects unrelated to that person's own work.

That breadth of access was linked, in the Reuters account, to OpenAI's use of a single large code repository, known in the industry as a monorepo. The sources said the repository was dense and complicated, which may have deterred some would-be attackers in the past. Now, they warned, AI agents can sift through that amount of material far more quickly.

Reuters also reported that OpenAI had not always carried out regular security reviews of internal tools, such as the sandboxes where researchers stress-test new models. Three people familiar with the company's practices said employees often modified those tools for their work, but those changes were not always tested for security problems, leaving them potentially vulnerable to attacks from outsiders.

OpenAI pushed back on parts of that picture. The company told Reuters that much of its most sensitive work is held in separate code repositories with strict access controls, and that the monorepo does not contain all of its code or critical details of its models, including the numerical values that determine how a model behaves, known as model weights. OpenAI also said it now conducts regular security reviews of its internal tooling, under chief information security officer Dane Stuckey, who joined in October 2024.

Background

The Reuters report connects the internal warnings to a series of incidents that have unfolded over the past year. On October 13, 2026, the US Justice Department said federal agents had arrested a Saudi national and charged him with installing a hidden audio recording device, concealed in toilet paper, in an executive conference room. Reuters identified the man as Abdulrahman Alfozan, 35, who was an OpenAI engineer at the time.

According to a court filing cited by Reuters, Alfozan also used an AI tool meant for OpenAI employees to find internal documentation about executives visiting Saudi Arabia and about which OpenAI officials were most likely to attend meetings with Saudi ministers. The filing said Alfozan was in regular contact with a Saudi royal who tried to recruit him for a new AI venture launching in Saudi Arabia. Lawyers for Alfozan did not respond to a request for comment, Reuters reported.

Earlier concerns, as described by the sources, date back further. In 2023, Leopold Aschenbrenner, then an OpenAI researcher, wrote an internal memo titled “OpenAI is China’s leading AI lab” about his concerns that China could infiltrate OpenAI and study its technology, three people familiar with the matter told Reuters. Aschenbrenner shared that memo with some board members later that year after an outsider hacked into OpenAI and took details about its GPT-4 model. OpenAI said the hacker accessed only Slack channels available to the whole company.

Security concerns flared again in the summer of 2026, Reuters reported, when OpenAI agents being tested hacked into the AI company Hugging Face, an incident that prompted scrutiny from state, local and federal officials, including the White House. Separately, cybersecurity firm Hacktron AI said three of its researchers used a rival AI model from Anthropic in July to break into ChatGPT accounts belonging to multiple OpenAI employees and gain access to some internal OpenAI systems, including its monorepo. OpenAI spokesperson Elana Widmann told Reuters the company thanked the researchers for sharing their findings and took steps to minimise the risk of similar attacks in the future.

Illustration: A shared code repository imagined as a vault, where broad access and layered controls sit in tension.

Why It Matters

The tension described by Reuters goes to the heart of how AI labs work. The field grew out of academia, where machine-learning scientists openly discussed their research, and OpenAI and Anthropic have tried to maintain a more open research culture to attract employees and encourage creativity. That openness, however, sits uneasily alongside the new status of frontier AI as a strategic asset for the United States, with pressure on American companies to protect their technology from foreign rivals.

Will DePue, a former OpenAI researcher who resigned from the company in April 2026, captured that tension in a post on X cited by Reuters: with competition so tense, AI leaders “can't sacrifice velocity for security. It is more important to them to move fast than to be secure, even from our greatest national adversaries.”

Reuters also noted a practical trade-off. Adding security layers can delay advances in AI technology, and could complicate plans by OpenAI and Anthropic for massive initial public offerings that would make them among the most valuable companies in the world. At the same time, failing to harden internal systems risks leaks, theft or manipulation of the very technology that gives those companies their value.

OpenAI is not the only AI company facing such problems, Reuters emphasised. In January 2026, a federal jury in San Francisco convicted a former Google employee who was arrested in March 2024 on charges of stealing thousands of pages of confidential information about Google's AI technology on behalf of China. In late October 2026, cybersecurity startup Cathedral Systems said it hacked into Anthropic's systems and gained access to Slack channels where employees discussed sensitive model details. An Anthropic spokesperson said the data accessed included some business contact information for customers and partners as well as a set of basic user account records.

For users, businesses and governments that rely on AI systems, the internal security of the labs building them is not an abstract concern. If model details, internal tools or employee accounts can be reached by outsiders, the risks range from stolen intellectual property to compromised accounts and, in the most sensitive cases, exposure of research that could be repurposed by hostile actors.

What Happens Next

OpenAI has begun to shift strategy, according to two people familiar with the matter cited by Reuters. Thibault Sottiaux, a senior executive overseeing product development who goes by “Tibo,” has recently taken charge of an internal initiative to secure OpenAI's internal systems. In August 2026, Sottiaux reassigned a quarter of OpenAI's production engineers to improve its security infrastructure, one of the people said.

After Reuters sought comment for its story, OpenAI said on Monday, November 2, 2026, in a blog post that it is planning its “next internal security sprint” to beef up its defences. Widmann said: “Protecting our technology, data, and people is a responsibility we take seriously and a matter of national importance. The actions we’ve taken recently show that program doing its job: identifying risks and responding decisively. We’ll keep strengthening our defenses as technology and threats evolve.”

In a separate appointment reported by Reuters on November 2, 2026, OpenAI named Dan Driscoll, a former US Secretary of the Army, as senior counselor for national security modernization. Driscoll will advise OpenAI's leadership and work with external partners on how advanced AI can enhance national security and cybersecurity, and help internal teams understand the operational needs of global national security organisations. “I'm joining OpenAI to help the United States and our allies put this technology to work responsibly, for the people who keep us safe,” Driscoll said in a statement cited by Reuters.

What to watch next is whether the security sprint translates into visible changes: tighter access controls, more frequent reviews of internal tools, and clearer separation between sensitive projects. Also worth watching is how regulators, investors and rival labs respond, and whether the debate over openness versus security reshapes hiring, research publication and the pace at which new models are released.

Illustration: A security team reinforcing digital defences around an AI system during an internal security sprint.

Sponsored: explore today's featured sponsored offers

FAQ

What did Reuters report about OpenAI's internal security?

Reuters reported, citing five people familiar with the matter, that OpenAI employees warned leaders for years about lax internal security, including broad access to code through a shared repository and internal tools that were not always reviewed for security problems.

What triggered the renewed attention in October 2026?

OpenAI disclosed in October 2026 that a recording device had been found in an executive conference room. The Justice Department said on October 13 that a then-OpenAI engineer, Abdulrahman Alfozan, had been arrested and charged over the device.

How did OpenAI respond to the Reuters report?

OpenAI spokesperson Elana Widmann said the company invests heavily in security against insiders, cybercriminals and nation-states, that sensitive work is held in separate repositories with strict controls, and that it now conducts regular security reviews of internal tooling.

What is a monorepo, and why does it matter here?

A monorepo is a single large code repository. Reuters reported that sources worried broad access to OpenAI's monorepo could expose details of many projects, while OpenAI said its most sensitive work and model weights are not all held there.

What is OpenAI doing next on security?

According to Reuters, senior executive Thibault Sottiaux is leading an internal security initiative, a quarter of production engineers were reassigned in August 2026 to improve security infrastructure, and OpenAI has announced a “next internal security sprint.” It has also appointed former Army Secretary Dan Driscoll to a national security modernization role.

Sources

Reuters, “OpenAI employees warned about operational security lapses for years, sources say,” November 3, 2026. Reuters, “OpenAI taps former US army secretary Dan Driscoll for national security role,” November 3, 2026. Reuters reporting on the October 2026 recording-device disclosure and the October 13, 2026 Justice Department announcement, as cited in the November 3, 2026 report.