Why are so many senior leaders leaving OpenAI right before its IPO?

OpenAI is arguably the most important company in technology right now, and its leadership is thinning out fast. Across 2026 a remarkable number of senior figures have walked, including several of the people responsible for making its systems safe. It is happening at an awkward moment: the company filed confidentially for an initial public offering in June and is being talked about as one of the largest listings in history. No single departure proves anything is wrong. Together, they have investors and insiders asking what the wave of departures says about OpenAI on the eve of a possible debut.

Update, September 2026

The picture has moved since this published, in one direction. The count has grown: what was "at least nine" in mid-August had, by late August, reached about a dozen by several reputable tallies. One genuinely new name joined the list, Chris Malone, OpenAI's head of data centres, whose exit was reported on 25 August as the company reorganised its infrastructure group under Sachin Katti. The IPO timeline also came into sharper, and later, focus: at an all-hands on 19 August, chief financial officer Sarah Friar told staff OpenAI "will be a public company in 2027," sooner only if the business keeps inflecting, and the confidential S-1 was still not public as of this update. The betting market repriced to match: Polymarket's implied odds of a 2026 listing, around one in five when this piece published, had fallen to roughly 8% by 10 September, with bettors now favouring 2027. The valuation ($852 billion at August's secondary sale) and the $40 billion run rate were unchanged. In short, the story the piece described, a concentrated leadership thinning against a slow march to the public markets, has if anything deepened.

Who has actually left?

The list, by name, is long for a single year. Confirmed or reported 2026 departures include:

• Brad Lightcap, a longtime deputy and former chief operating officer who had moved to lead special projects; he joined in 2018 and announced in August that he is leaving to start something new.
• Denise Dresser, chief revenue officer, out after about eight months and announced within days of Lightcap's own exit; she is being replaced by Dali Rajic, formerly president and chief operating officer of the security firm Wiz.
• Fidji Simo, widely regarded as the company's No. 2, who stepped down in July, reportedly for health reasons.
• Bill Peebles, who led the Sora video effort, along with Kevin Weil and Srinivas Narayanan, who left earlier in the year.

Then there is the group that has drawn the most attention:

• Chloé Bakalar, described as OpenAI's only dedicated AI ethicist, who left in July. By multiple accounts her exit was not announced and her role has not been filled.
• Johannes Heidecke, who led OpenAI's Safety Systems work, who announced his departure in July.
• Joshua Achiam, head of mission alignment turned "chief futurist," reported to be leaving as well, and described in coverage as at least the fifth senior safety-focused leader to depart in about two years.

The company's mission-alignment team, formed in 2024, was disbanded earlier in 2026. By one accounting, first reported by former staff, roughly half of the safety researchers OpenAI had in 2023 and 2024 have since left.

Why does it matter that the safety people are leaving?

Because OpenAI's entire public case rests on building systems more powerful than any before, safely. When several of the people hired to do the "safely" part leave in a cluster, and one of the vacated roles is simply left empty, it raises a question about how much weight that half of the mission still carries inside the company. The counter-argument is that safety work is spread across many teams, not any one person or role. But the optics, a thinning safety bench just as capabilities and commercial pressure climb, are exactly what critics have warned about for years. (For the flip side of that worry, see our look at AI models behaving unexpectedly on the live internet.)

What does this have to do with the IPO?

Timing. OpenAI filed confidentially for an IPO in June and is being discussed at a valuation reported around $850 billion, with some framing it as a march towards a trillion-dollar listing. A public offering rewards exactly what a run of senior departures can unsettle: stability, a clear bench, and investor confidence in the long-term plan. "It is just not that typical to have so many executives depart before their long awaited IPO," Jeff Park of ParaFi Capital told reporters. One AI startup founder was blunter, calling the wave a "huge red flag." (OpenAI is not the only lab eyeing the public markets; see Anthropic's own move towards an IPO.)

The market is also sceptical that a listing is imminent. On the prediction market Polymarket, the implied odds of OpenAI going public by 31 December 2026 sat at around one in five when this piece published (a figure that, by early September, had fallen to roughly 8% as OpenAI's own guidance shifted to 2027, see the update above). That figure says more about timing than about the departures: a company that only filed confidentially in June, and still faces regulatory review and a roadshow, rarely lists within months. It is a betting-market signal, not a company statement.

Is there an innocent explanation?

Partly, yes, and it deserves saying. Some of these exits have ordinary causes: Lightcap is starting a venture, Simo cited health, and senior people leaving a fast-growing company after years to cash out or build their own thing is common in Silicon Valley. OpenAI is not struggling financially. It is on a reported $40 billion annualised revenue run rate, roughly double its run rate at the end of 2025, and a $7 billion secondary share sale in August gave staff a way to realise gains without waiting for an IPO. Naming an immediate replacement for Dresser suggests a functioning succession process, not disarray.

The honest read is that no single departure is damning, and the company is clearly growing. What makes this a story is the concentration and the composition: a lot of senior people in a short window, weighted towards the safety and mission side, at the precise moment the stakes, and the scrutiny, are highest.

What OpenAI says

Leaders on the way out have largely framed their exits positively; Lightcap said "mission success feels within sight." OpenAI has kept announcing new hires and promotions alongside the departures, and points to its growth and product momentum. Chief executive Sam Altman has kept the focus on the company's scale and its IPO ambitions. OpenAI has not described the departures as a crisis, and there is no evidence they are coordinated.

Frequently asked questions

How many executives have left OpenAI in 2026?

At least nine senior leaders by name had departed or announced departures by the time this piece published in mid-August; by late August, reputable tallies had reached about a dozen. They include the former chief operating officer, the chief revenue officer, the company's No. 2, and several safety and ethics leaders. The pace picked up in August, with Brad Lightcap and Denise Dresser leaving in the same week, and OpenAI's head of data centres, Chris Malone, following later in the month.

Are the departures about AI safety?

Not provably. Several leavers gave unrelated reasons, such as a new venture or health. But a notable share of those leaving worked on safety, ethics or mission alignment, and by one accounting about half of OpenAI's 2023-24 safety researchers have since gone, which is why the safety angle is drawing the most concern.

Is OpenAI in financial trouble?

No. OpenAI is on a reported annualised revenue run rate of more than $40 billion, roughly double its run rate at the end of 2025, and completed a $7 billion secondary share sale in August. The concern is about leadership stability and the safety bench, not revenue.

When is OpenAI's IPO?

Unconfirmed, and now pointing to 2027. OpenAI filed confidentially in June 2026 but has not set a date, and in August its chief financial officer, Sarah Friar, told staff the company would be public in 2027 unless growth accelerates. Prediction-market bettors on Polymarket, who put the odds of a 2026 listing at around one in five when this piece published, had marked them down to roughly 8% by early September, favouring 2027 instead.