When Sam Altman first instructed her that he’d by no means let OpenAI go company, that what he and his colleagues have been constructing was too highly effective to be pushed by traders, Catherine Bracy kind of believed him.

The dialog occurred in 2022, when Bracy, CEO and founding father of the social mobility-focused nonprofit TechEquity, was interviewing Altman for a book she was writing concerning the risks of enterprise capital. It was earlier than Altman’s mysterious firing and unfiring a 12 months later, after which he largely stopped responding to Bracy’s texts.

And ever since then, OpenAI — which was initially based as a nonprofit in 2015 to “advance digital intelligence in the way that is most likely to benefit humanity as a whole, unconstrained by a need to generate financial return” — has been publicly making an attempt to flee the confines of its charitable roots. At the moment, OpenAI accommodates each a company arm centered on constructing and promoting AI and a nonprofit arm with a acknowledged mission of making certain that AI advantages folks.

Throughout the controversial strategy of making an attempt to fully sever the 2 in 2024, OpenAI lost about half of its AI security staffers and far of its senior leadership. That was adopted by an intensified scrutiny from state attorneys general, nonprofit legal experts, competitor companies, effective altruists, Nobel Prize winners, huge swaths of California’s philanthropic community, and one among its unique funders, Elon Musk. Totally different sides had totally different pursuits, however the general argument was that shifting to a for-profit mannequin would create a fiduciary responsibility to traders that might inherently conflict with its unique mission of security and public profit.

Nonetheless, OpenAI did lastly strike a contortive restructuring deal final October. Primarily, the for-profit arm turned what is named a public profit company (PBC), referred to as the OpenAI Group. The unique nonprofit turned the OpenAI Basis, which has a 26 percent stake currently worth $180 billion within the PBC, plus a sliver of unique authorized management over sure main selections.

One impact of the transition was that it primarily required OpenAI to place a quantity on what it owed the general public for changing what had been a mission for all humanity into one thing that almost all immediately advantages the corporate’s traders. The ensuing stake of the OpenAI Basis is sufficiently big to immediately make it one of many wealthiest charities within the nation, or in OpenAI’s words, the “best-equipped nonprofit the world has ever seen.” On paper, at the least, the muse is now considerably richer than the entire country of Luxembourg. Even the Gates Basis has solely $77.6 billion in property, lower than half of what the OpenAI Basis can draw from, although it’s necessary to notice that many of the wealth of the OpenAI Basis is locked in pretty illiquid shares inside the nonetheless personal firm, which limits how shortly any cash will be given away.

Nonetheless, its sheer dimension implies that the OpenAI Basis stands to finally be a transformative presence on the philanthropic stage, a technique or one other. However whereas OpenAI says the muse will finally give out many billions of {dollars} in philanthropy to make sure that “synthetic normal intelligence advantages all of humanity,” it’s unsure {that a} socially useful philanthropy can exist facet by facet with an organization that’s preventing an existential battle over who will dominate the AI trade.

“The unstated reality right here is that they’re by no means going to decide that’s dangerous for the corporate,” Bracy stated. “These two entities can’t dwell underneath the identical roof” the place “the mission is in management.” (Disclosure: Vox Media is one among a number of publishers which have signed partnership agreements with OpenAI. Our reporting stays editorially impartial.)

The muse’s first items got here within the type of $40.5 million in no-strings-attached grants to over 200 group nonprofits, like church buildings, meals banks, and afterschool packages. Notably, most grantees had little to no connection to AI or expertise — and simply as notably, a number of of those early grantees simply so occur to be members of EyesOnOpenAI, a coalition of California nonprofits vital of OpenAI’s privatization that shaped in 2025.

However there are indicators the muse will quickly pivot into grantmaking that’s extra clearly related to the corporate’s unique constitution, which aimed to make sure that the advantages of AI are broadly distributed whereas additionally prioritizing long-term security within the expertise’s improvement. On Feb. 19, OpenAI — the corporate, not the muse — introduced a $7.5 million grant along side Microsoft, Anthropic, Amazon, and different main tech firms for a new, international project aimed toward researching the best way to make AI programs safer.

The true questions across the OpenAI Basis have much less to do with how a lot it’s giving and to whom than whether or not it’s really in a position to perform its contractual oversight position. In idea, the muse needs to be making certain that OpenAI is the standard-bearer for moral decision-making on the frontier of AI improvement. That might be a novel contribution to the sector — and an embodiment of OpenAI’s unique mission — that no quantity of grantmaking may substitute. But, a sequence of troubling latest selections by the corporate hardly appears to bear out that imaginative and prescient.

OpenAI has begun its new company journey by debuting ads on its free tier service, firing an executive who raised security issues a couple of soon-to-come NSFW mode for ChatGPT on costs of sexual discrimination in opposition to a male colleague, and burning cash whereas its president funnels tens of millions of {dollars} into Donald Trump’s super PAC. OpenAI President Greg Brockman has additionally teamed up with the personal fairness agency Andreessen Horowitz and Palantir’s co-founders to fund a $125 million super PAC aimed toward selling AI-friendly insurance policies. Together with Google, xAI, and Anthropic, OpenAI has additionally come underneath scrutiny in latest weeks for its defense contracts with the Pentagon.

When OpenAI succeeded in its marketing campaign to cede its foundational new expertise from nonprofit management, it opened the door for a lot of of those selections. Even $180 billion in charity won’t be sufficient to make up for the distinction.

How OpenAI shed its nonprofit pores and skin

Company charity is ubiquitous within the tech world, particularly among the many largest gamers. Microsoft plans to donate $4 billion in cash and AI cloud technology to colleges and nonprofits by 2030. Google offers away some $100 million yearly, usually to organizations centered on synthetic intelligence and expertise.

However from the start, OpenAI was totally different. Somewhat than getting cash and giving a few of it to charity, OpenAI was the charity. It was based as a nonprofit analysis lab with about $1 billion in start-up donations, largely from tech titans like Altman, Brockman, and Elon Musk.

There are some structural benefits to being a charity. You possibly can’t settle for investments, however you possibly can settle for donations and also you don’t should pay most taxes. What’s extra, in these early days, OpenAI’s acknowledged mission — to construct secure AI with out the pressures of monetary incentive — gave it a significant enhance when it got here to recruitment for rarified expertise. Machine studying prodigy Ilya Sutskever told Wired in 2016 that he selected to go away Google to change into OpenAI’s chief scientist “to a really giant extent, due to its mission.”

However there have been limits to being a totally nonprofit entity. In pursuit of financing amid the rising computing prices of cutting-edge AI, OpenAI created its capped-profit subsidiary in 2019 to handle a brand new $1 billion funding from Microsoft. Three years later, ChatGPT took the world by storm. Sutskever, and different members of OpenAI’s board, tried and ultimately failed to oust Altman amid accusations of dishonesty in 2023. (Altman denied these accusations.) In 2024 — one 12 months after Sutskever and different members of OpenAI’s board tried and finally did not oust Altman amid accusations of dishonesty — the group announced its intention to go totally company and splinter off the nonprofit into its personal totally impartial entity.

The transition to for-profit “simply didn’t odor proper,” stated Orson Aguilar, head of LatinoProsperity, an financial justice nonprofit and Bracy’s co-leader at EyesOnOpenAI. He wasn’t alone: By early 2025, a dozen former OpenAI staff filed an amicus brief aimed toward stopping the conversion as a result of it could “basically violate its mission.” And greater than 60 nonprofit, philanthropy, and labor leaders, lots of them primarily based in OpenAI’s dwelling state of California, agreed that the try and privatize felt unfair given the extent to which the corporate benefited from its tax-free standing throughout its early improvement.

However California historical past did present some hope that the general public may at the least get some significant profit from the transition. Again within the Nineteen Nineties, California’s department of the well being insurer Blue Cross Blue Defend — then a nonprofit referred to as Blue Cross of California — determined to denationalise. After some haggling with state regulators, the corporate agreed to forfeit all of its property, worth $3.2 billion, to a pair of impartial nonprofits in alternate for going personal. The consequence was the California Endowment, which is now the state’s largest well being basis.

Many nonprofit leaders in California hoped that OpenAI, which is headquartered within the state, would strike an analogous deal, ceding a majority of its property to a totally impartial nonprofit. And people property have been and are huge.

Gary Mendoza, a former state official who oversaw the Blue Cross deal, estimated the OpenAI nonprofit’s rightful property at over $250 billion, or half the corporate’s $500 billion value. “Something in need of 50 %,” he instructed the San Francisco Examiner final 12 months, “is a missed alternative.” And past cash for the general public, assuming the nonprofit saved its shares, it could add as much as sufficient affect to actually form OpenAI’s company decision-making at a key second for the way forward for synthetic intelligence.

Provided that the OpenAI Basis ended up with little greater than 1 / 4 of the ultimate firm, that is clearly not what occurred. However EyesOnOpenAI’s years-long lobbying effort was not a complete bust. The criticism proved highly effective sufficient that last May, OpenAI was compelled to surrender on an preliminary plan to restructure away its nonprofit property into a brand new group wholly disconnected from OpenAI, which might have left the nonprofit with no authorized management over the for-profit arm.

On paper, the new deal consists of some significant concessions. It contractually requires the nonprofit mission to come back first on security and safety points, with no regard to shareholder pursuits. The memorandum additionally calls on OpenAI to “mitigate risks to teens” particularly. It made the muse the controlling shareholder of the company, affording it the appropriate to nominate company administrators and oversee vital selections like a sale.

If OpenAI abided by all of its phrases and finally began giving freely billions of {dollars} of philanthropy annually, then the world — or at the least California, the place lots of OpenAI’s grants have been concentrated — may stand to vastly profit from it.

Random acts of company kindness

And this brings us to the $40.5 million that OpenAI gave to over 200 nonprofits towards the tip of final 12 months.

Many of those charities utilized to the grant with refined concepts round the best way to assist their communities combine or adapt to AI, although they will finally use the grants nonetheless they see match. Among them were public libraries, Boys and Ladies Golf equipment, church buildings, meals banks, and authorized support nonprofits. Coming at a second when nearly all of the nation’s nonprofits face existential funding cuts, “it was simply the right timing,” stated Thomas Howard Jr, head of Kidznotes, a North Carolina nonprofit centered on music schooling that acquired $45,000 in OpenAI’s first spherical of grants.

So civil society’s battle over the OpenAI transition received at the least sufficient concessions to assist these worthy organizations and retain some semblance of nonprofit management over among the for-profit’s actions. So why achieve this many individuals within the philanthropic group stay so unfavorable concerning the basis?

“I’m all for nonprofits getting cash,” stated Bracy, the top of TechEquity. “I don’t begrudge any organizations that took the cash, however I don’t assume it’s some indication that OpenAI resides as much as the mission of the nonprofit.”

$40.5 million, in fact, is barely 0.02 % of the OpenAI Basis’s on-paper $180 billion windfall. How the muse will finally spend the opposite 99.98 % stays to be seen, although the muse has stated that at the least $25 billion will finally go to scientific analysis and what it’s calling “technical options for AI resilience.” The corporate plans to announce a second wave of grants directed at organizations utilizing AI to work throughout points like well being within the coming months.

“We’re doing the necessary work of participating with specialists, studying from communities, and shaping a viewpoint of the place Basis investments could make the best distinction,” the OpenAI Basis’s board of administrators stated in response to a request for readability on the place future funding will go. “We look ahead to sharing extra quickly.”

However up to now, critics stay skeptical. OpenAI has accomplished little to show that its newfound philanthropy is extra than simply “a smoke and mirrors present,” argued one member of the Coalition for AI Nonprofit Integrity (CANI) — a coalition composed largely of AI insiders, together with former OpenAI staff, furiously against the restructuring. He spoke on the situation of anonymity as a result of he feared retaliation from OpenAI, which has accused CANI of being a front funded by Musk. (CANI has denied receiving any such funds — although not for lack of trying. In the event you scroll to the underside of OpenTheft, a web site created by CANI, you’ll discover a direct plea to Musk for donations.)

Whereas a spokesperson for OpenAI stated that the muse is within the strategy of constructing a devoted staff, and has sought the enter of each nonprofit leaders and specialists in how society can adapt to AI, the corporate has but to make any main staffing bulletins for its grantmaking arm. For now, except Zico Kolter, the top of the nonprofit’s security committee, the muse board nonetheless shares the identical members as the company board, together with CEO Sam Altman. The concept is that these board members can placed on totally different hats when assembly about nonprofit versus company priorities, asserting the muse’s oversight when wanted. Nevertheless it has created the looks of a battle of curiosity.

When requested for mechanisms and examples for the way the muse has responded to conditions the place its mission conflicts with shareholder pursuits, given the overlapping board membership, the spokesperson stated that OpenAI has conflict-of-interest insurance policies and governance procedures in place to make sure its administrators solely think about the mission after they meet, as they commonly do, about nonprofit points.

The corporate additionally stated the muse board continuously workout routines its oversight position, together with for all new main product releases, like the discharge of GPT‑5.3‑Codex, a complicated agentic coding mannequin, final month. The AI watchdog group the Midas Project, a frequent thorn in OpenAI’s side, accused the corporate of violating security requirements, an allegation that OpenAI fervently denied.

In any case, because the OpenAI Basis isn’t a separate entity with its personal impartial board, some critics have in contrast it to different feel-good company social duty ventures, just like the McDonald’s Ronald McDonald House, Walmart’s wholesome meals program, and Home Depot’s work with veterans.

Company social duty has its place, and it might probably do actual good. However Bracy believes that primarily based on the OpenAI Basis’s structuring and the way they’ve carried out their grantmaking up to now, it would in all probability by no means fund something “they see as a menace to the expansion of the corporate,” stated Bracy, even though the necessity for guardrails on unrestricted AI improvement featured prominently within the firm’s unique mission. “They’re going to do what’s greatest for the underside line of the for-profit.”

Critics like Bracy additionally doubt the OpenAI Basis’s different essential prerogative, which is to manipulate all security and ethics-related points for the broader group, together with the duty to overview new merchandise.

Whereas the nonprofit and its mission do legally retain management over the OpenAI company — significantly with regards to issues of safety — which will add as much as little, provided that the OpenAI Basis doesn’t appear to be an independently ruled basis. It’s not, in truth, even technically a basis, however a public charity, which suggests it isn’t required to pay out a sure share of its property annually underneath IRS necessities.

And whereas the nonprofit retains vital oversight powers on paper — together with the authority to halt AI releases it deems unsafe — in observe, critics say, it’s unclear whether or not it could ever use them.

More and more, OpenAI has additionally been wading into political lobbying efforts that appear at odds with its mission to advertise long-term security in AI improvement. When California lawmakers have been debating SB 53, a legislation requiring transparency studies from main AI firms, OpenAI lobbied in opposition to it. And the corporate has come underneath intense scrutiny in latest weeks for its contract with the Pentagon, which has blacklisted its rival firm Anthropic for elevating moral issues about using its expertise.

Why the battle isn’t over

OpenAI’s new company association may be very, very new. It’s nonetheless attainable that OpenAI’s grantmaking arm actually does workers up, and the nonprofit builds an impartial board that has the facility to implement exhausting moral selections for the corporate, even when it hurts traders’ returns.

“They’ve a whole lot of freedom to proceed to do good,” stated Tyler Johnston, government director of the Midas Project, however that might require them to “really shake issues up” and “present that they’ve created the scaffolding that can allow them to actualize their mission.”

However up to now, “there’s nothing I’ve seen that provides me reassurance that they’ll catch the necessary issues of safety after they come up,” he stated. “Or that they’ll be doing an intensive investigation of the grantmaking alternatives.”

If OpenAI doesn’t abide by the phrases of its new contract — if the corporate, for instance, tries to thwart an try and roll again a harmful new software — then California’s legal professional normal does have the facility to demand solutions from the corporate, and in idea, revisit the settlement’s phrases.

Past the settlement, there are a couple of fairly public means by which OpenAI’s former lovers, skeptics, and nemeses are nonetheless making an attempt to press rewind on the restructuring.

Chief amongst them is Elon Musk, OpenAI’s most outstanding unique donor and co-founder. In between buying and selling embarrassing jabs with Altman on X, Musk took OpenAI to court docket final 12 months over claims that he was “assiduously manipulated” into donating tens of tens of millions of {dollars} to a nonprofit analysis lab that changed into an “opaque internet of for-profit OpenAI associates.”

A decide has discovered sufficient trigger for the case to proceed to trial this April. Musk is suing for as much as $134 billion in damages, although OpenAI has told its investors that it believes it could solely be on the hook for Musk’s $38 billion in unique donations. OpenAI, for its half, has accused Musk of an “unlawful campaign of harassment.”

In the meantime, CANI remains to be holding out hope that it might probably persuade the folks of California to vote for a hyperspecific poll measure, the California Charitable Assets Protection Act, which may reverse the choice to permit OpenAI — or every other “organizations creating transformative applied sciences” — to go company.

“They’re chopping corners on security due to the race to synthetic normal intelligence that they only need to win,” stated the member of CANI. “As a substitute of a automobile to serve humanity, it’s change into a automobile to serve one particular person and some of his pals and traders.”

So perhaps the battle over OpenAI’s restructuring isn’t utterly over — but it surely’s in all probability on its final legs. And in the event that they proceed on the identical path, it’s unlikely that the general public will ever actually profit in the best way they should, given the charitable advantages OpenAI loved in its early days. On the very least, $40.5 million is simply not going to chop it. Even $180 billion may fall far quick.

“I feel it’s them saying, ‘Pay attention, I dare you to implement this,’” stated Bracy, who believes OpenAI is “banking on the truth that they’re value virtually a trillion {dollars}, and so they have countless sources — and the state of California doesn’t.”