The proposal landed days after the request to delay. Sam Altman has offered the U.S. government a 5% stake in OpenAI as a “national interest safeguard,” in exchange for a fast track to approval for the company’s next flagship model, GPT-5.6, according to the Financial Times. The offer came just days after the White House asked OpenAI to postpone the model’s release.
The mechanics are still being worked out. People familiar with the proposal said the stake would take the form of equity in OpenAI, with the structure, voting rights and board representation still under discussion. At the company’s current valuation, 5% is worth billions of dollars, a price Mr. Altman has signaled he is willing to pay for regulatory certainty.
The context is a standoff that has been building for weeks. OpenAI says GPT-5.6 is ready, tested and safer than its predecessors; administration officials, citing national security concerns raised by intelligence agencies, have asked for more time to review it. The two sides have been negotiating behind closed doors, and the equity offer is, according to people familiar with the talks, Mr. Altman’s attempt to convert an adversarial process into a partnership.
The proposal has a precedent of sorts. The U.S. government has taken equity in private companies before, most visibly when the Treasury received warrants in automakers and banks during the 2009 financial crisis. But it has never held shares in an artificial intelligence company, and the stakes are different: a stake in OpenAI would give Washington a continuing say in how a leading AI lab is governed, not just a claim on future profits.
Supporters of the arrangement say it is a rational trade. OpenAI gets a predictable regulatory path, and the government gets a permanent seat at the table, with access to the company’s finances, its decisions and its technology road map. Mr. Altman has described the offer, according to people familiar with his thinking, as a way to give the government a stake in OpenAI’s success rather than a reason to fear it.
Critics see a different problem. A government shareholding in a dominant AI company, they argue, blurs the line between regulation and ownership, and could make it harder for Washington to police the industry it partly owns. Some members of Congress have already raised questions about the terms, and antitrust experts have noted that a government stake does not address the underlying concerns about concentration in the AI market.
The proposal also has implications for OpenAI’s existing investors, who would see their ownership diluted to make room for the government. People familiar with the shareholder discussions said the terms would need to be structured carefully, with the government’s stake likely priced at the valuation of the latest funding round. Microsoft, which owns a large share of OpenAI, and the venture funds that have backed the company would all be affected, though none has publicly objected.
The offer also raises governance questions that OpenAI has not answered. Would the government’s shares carry voting rights? Would a Treasury official sit on the board? How would the stake be valued in a future IPO, and would other investors be diluted to make room for it? People familiar with the discussions said these questions are exactly what the two sides are now negotiating.
The timing gives Mr. Altman room to negotiate. OpenAI has already released a preview version of GPT-5.6 under the name GPT-5.6 Sol, a move widely read as a signal that the technology is finished and the delay is political. The company has been telling customers that the full release will come as soon as the review concludes, and the equity offer is part of that campaign: make the approval process a matter of shared interest rather than an obstacle.
Analysts were divided on the proposal’s merits. Some described it as shrewd, noting that a 5% stake is a small price for removing the biggest risk in OpenAI’s business: an unpredictable government. Others said it sets a dangerous pattern, in which every AI company with a controversial product negotiates by offering Washington a piece of itself. “If this works, every lab will try it,” one technology policy analyst said. “The question is whether the government should be in the business of taking equity in exchange for doing its job.”
For now, the offer is on the table and the review continues. The White House has not responded publicly, and officials have said only that the safety review will proceed on its own schedule. Mr. Altman, meanwhile, has made his position clear: 5% of the company, he has effectively said, is a fair price for the one thing OpenAI cannot buy on the open market, which is certainty.
If the deal goes through, it will be a first: the U.S. government as a direct shareholder in an artificial intelligence company. If it fails, the standoff over GPT-5.6 continues, and the preview release keeps carrying the weight of the negotiation. Either way, Mr. Altman has changed the terms of the conversation, from whether the government should slow OpenAI down to what it would take to let it go.
