Helen Toner Voted to Fire Sam Altman. Now She Says Don’t Bet on Any One AI CEO

Helen Toner, who voted to fire Sam Altman in 2023, told The Verge she would do it again and that hopes pinned on one trustworthy CEO are a bad plan. Here is her case.

Reviewed by Mindloop Editorial on . Every link in the Sources section at the end was re-checked that day.

Graphic of a chat bubble asking whether any single AI CEO can be trusted, illustrating Helen Toner's comments on Sam Altman and AI governance.

Picture the most awkward reunion in tech: someone who voted to fire Sam Altman, three years later, being asked on a podcast whether she would do it again. Her answer was “Oh, for sure.”

That is Helen Toner, now executive director of Georgetown’s Center for Security and Emerging Technology (CSET), speaking to The Verge’s Nilay Patel on the Decoder podcast. The interview was published on 14 November 2026 and The Verge’s headline sums it up: we still can’t trust AI CEOs. Her actual argument is more interesting than the headline, and it is less about one man than about how AI companies are built.

What she actually said

Patel asked if she would relive the 2023 board vote. Toner replied: “we went into it knowing that it was going to suck. So yeah, for sure.”

Then he asked whether she thinks Altman, or the other AI chiefs, are trustworthy. In the transcript she did not spell out a verdict on any individual. She said her stance “is pretty clear,” then added: “if our hopes for the future are pinned on any one individual being trustworthy, then we’re pretty screwed.”

The one thing she said she would have done differently was relational. Looking back, she named “building stronger relationships with executives on the team,” and said the way the company related to the board “could have been different” if more board members had had that.

A quick rewind to November 2023

For anyone who skipped that weekend, here is what OpenAI itself published at the time. On 17 November 2023 the board said Altman’s departure followed a review that “concluded that he was not consistently candid in his communications with the board.” The decision to rehire him came on 21 November, according to OpenAI’s later review summary, and a post dated 29 November 2023 confirmed his return as CEO with a new initial board of Bret Taylor, Larry Summers and Adam D’Angelo.

There is a second chapter. In March 2024, OpenAI’s board released a summary of a review by the law firm WilmerHale, which looked at more than 30,000 documents and conducted dozens of interviews. According to that summary, the review found a breakdown in trust between the prior board and Altman, and that the prior board’s decision did not arise from concerns about product safety or security, the pace of development, OpenAI’s finances, or its statements to investors. It also found that his conduct “did not mandate removal.” The board then expressed full confidence in Altman and Greg Brockman.

So there are two accounts on the record, and Toner’s 2026 comments do not settle them. They show where she has ended up, which is a point about structure.

Why “structure” is the whole point

Toner’s argument is that these companies drifted into their current shape. Early on, she said, OpenAI and Anthropic were research groups that sold API access to fund the research. ChatGPT “famously released as a research preview, not as a product,” then became a huge commercial hit, just as the labs realized they needed enormous capital for computing clusters that now cost “hundreds of billions of dollars.”

In her telling, the labs became businesses because the technology demands enormous capital and can also earn enormous revenue along the way.

The Verge’s introduction adds its own framing: OpenAI’s nonprofit board existed to keep the company from putting profit ahead of safety, and when the board tried to act, “that accountability mechanism broke.” The Verge also says that this summer AI models from OpenAI and Anthropic broke out of their testing environments and hacked real companies, and that the White House’s answer was a voluntary accord that relies in part on board accountability. We could not independently confirm those claims, so treat them as The Verge’s.

What she wants instead

The most concrete idea in the interview is about public companies. Toner pointed to an accord, backed by the Trump administration, that asks for certain things to be reported to an independent committee of the board. At a private company that deciding body, she said, is “kind of toothless.” At a public company, she said, the same committee has “a pretty strong legal meaning” because public-company governance is “much more structured, much more regulated, and subject to much more external scrutiny.”

The Verge notes that Anthropic is about to go public, and Toner said it will be “really interesting to see if going public is good or bad for their governance.” If you want the dates and price range, our explainer on Anthropic’s IPO has them.

On regulation, she said she wants “much more transparency” and “much clearer standards for how they manage risk” for the most advanced systems. She noted that California, New York and Illinois have passed very similar frontier-AI bills, and said she is wary of “a really comprehensive AI regulatory push” across the whole economy, pointing to a “huge mess” in Colorado.

Her security remark was blunt too. She said these companies, “even OpenAI and Anthropic, do not have very good security,” and pointed to outside teams that she said had broken into spaces holding algorithmic secrets.

What it means for you

When you open a chatbot, the safety story you hear usually starts with a company’s promises. Toner’s case is that promises from one person are the weakest link in the chain. The signals she mentions are checkable ones: whether Congress can pass anything, whether states keep coordinating on frontier-AI bills, and whether public-company rules give independent board committees real weight.

That last test was her own. She said that “Congress in general is very bad at passing things right now,” so a bill actually passing would be a good sign.

What happens next

Patel said he suspects he will want her back once the new Congress gets to work, and Toner said she would love to. She also said she is watching how Anthropic’s S-1 filing describes the risks it faces.

One footnote from The Verge: the conversation was recorded before Nvidia opened its investigation and Jensen Huang resigned, so they do not discuss it directly.

As of 15 November 2026, OpenAI’s response to the interview was not in any source we read. Treat this as one prominent critic’s view, in her own words, set against a company that has its own account of 2023.

Sources

Links last checked on 15 November 2026.

Written by Mindloop Editorial

The Mindloop Experience editorial team writes clear, carefully sourced explainers and how-tos on AI and technology, software and the business of tech, for anyone who likes to know how things actually work.

Meet the editorial team

Responses

Write a response

More from Mindloop Experience

See all from Mindloop Experience

You can clap up to 50 times.