OpenAI is expanding its fight against organizations opposed to its for-profit transition, delivering a subpoena this week to the The Midas Project that accuses the AI watchdog of being funded by Elon Musk.
OpenAI has issued similar subpoenas in recent months to at least three other AI safety nonprofits as part of its legal battle with Musk, The Standard reported Tuesday. Musk, the cofounder of competitor xAI, sued OpenAI last year, accusing it of betraying its nonprofit mission by seeking to become a for-profit entity. OpenAI countersued and has been in a legal back-and-forth with the billionaire ever since.
But nonprofit groups like The Midas Project say they are getting caught in the crossfire between the tech titans — and are being falsely accused of being fronts for OpenAI’s adversaries.
“I think that they’re going after all of the nonprofits that have a substantial chance of impacting what their restructuring might look like,” said Tyler Johnston, the founder of The Midas Project, calling the subpoenas an “effective way to slow down nonprofits.”
“If I have to produce the documents, because I’m the only full-time staffer, it could be my new job,” he added.
In a statement, OpenAI lawyer Anne O’Leary said the company was just looking for transparency from its critics.
“Unless The Midas Project also has the Midas touch, when it comes to their financial backers someone is funding their attacks on OAI,” she said. “What we are asking for is for them to be transparent as to who is funding these efforts and what the financial interests of these funders are.”
The Midas Project largely monitors changes to the safety policies of major AI players, including xAI, Anthropic, and OpenAI. It is best known for the OpenAI Files, a 50-plus-page collection of documented concerns with the company’s governance practices and organizational culture, which the nonprofit published in June in collaboration with the Tech Oversight Project.
The Midas Project also helped organize a letter this summer from more than 100 prominent figures — including Nobel Prize winners, professors, and former OpenAI employees — demanding more transparency about the company’s transition plans.
Johnston told The Standard he was traveling last month when he received a text from his roommate in Oklahoma saying there was someone at the door trying to serve him a subpoena. After some back and forth, Johnston received the documents — actually containing two subpoenas — by email Monday.
The subpoenas — one for The Midas Project and one for Johnston personally — ask for any communications with Musk or affiliated entities around the formation of The Midas Project, any communications with Musk or Meta CEO Mark Zuckerberg about The Midas Project’s positions and activities, and any contributions made by Musk to the nonprofit.
(OpenAI’s lawyers allege Musk contacted Zuckerberg about collaborating on an attempted takeover bid he launched in February. Zuckerberg allegedly declined.)
The subpoenas, which are dated August 12, go beyond Musk and Meta. They also ask for the identities of all The Midas Project’s funders, as well as the dates and amounts of their donations, and any documents or communications the group possesses regarding the governance or organizational structure of OpenAI — essentially, all of its work around the restructuring issue.
Johnston said his group has never received funding from Musk or Meta, nor has he spoken to Musk about the project. He said he understands OpenAI’s speculation about a possible Musk involvement, but that the demands for all of his funders and all his work on the restructuring seem like a fishing expedition.
“The only thing that is relevant to their counterclaim here is, ‘Is this a secret Musk harassment campaign?’” he said. “If they were just asking that, I'd be fine. But no, they're actually taking the opportunity to try to stifle or expose advocacy efforts that are inconvenient to their interests as a company.”
O’Leary defended the company’s requests in her statement, saying: “All these groups, including the Midas Project, are very focused on transparency—except when it comes to who is funding their own efforts.” (Johnston said OpenAI had never asked about his organization’s funding before issuing the subpoena.)
OpenAI has also sent subpoenas to three other nonprofit groups opposing its for-profit transition in recent months: Encode, Legal Advocates for Safe Science and Technology, and the Coalition for AI Nonprofit Integrity. Representatives of all three organizations say they have never spoken with Musk or received any funding from him.
Encode did submit an amicus brief in Musk’s lawsuit supporting some of his arguments, and CANI worked on a bill earlier this year that would have blocked OpenAI’s transition to a for-profit entity. Johnston said he signed an early open letter from CANI supporting the bill and took two meetings with the group, but did not help them write the bill or lobby for it. He also said he was not aware of Musk’s lawsuit at the time Encode filed its brief.
Complicating OpenAI’s narrative, The Midas Project has been critical of Musk’s AI efforts as well. It published articles critical of his xAI, and just two weeks ago signed onto a letter to the Consumer Federation of America calling for an investigation into the company’s tolerance for nonconsensual deepfake images.
But Johnston’s history with a different billionaire may be what OpenAI is actually concerned about. Representatives for the company have recently pointed fingers at the billionaire backers of the effective altruism movement, including Facebook cofounder Dustin Moskovitz.
Moskovitz’s Open Philanthropy Foundation has poured hundreds of millions of dollars into think tanks and research groups pushing lawmakers to consider the existential risks of unchecked artificial intelligence. He is also an investor in Anthropic, a competitor to OpenAI, though he has since donated his shares in the company to a nonprofit not affiliated with Open Philanthropy.
Johnston said he took a grant of $35,000 from Open Philanthropy before officially founding The Midas Project, as part of his career transition into pursuing AI safety. He has also been tangentially involved in effective altruism, including through his animal rights work.
But The Midas Project has published articles critical of Anthropic, including a recent report titled “How Anthropic’s AI Safety Framework Misses the Mark.” And Johnston sits on the board of another nonprofit called PauseAI, which recently organized a protest outside Anthropic’s headquarters.
Johnston said he makes about $25,000 a year as the only full-time employee of his organization, and laughed at the implication that he has somehow been “bought out” by corporate interests.
“If Musk wants to buy me out, cut me a check for $10 million, and I will start saying pro-Musk things and move to a beach somewhere,” he joked. “But until that happens, I don’t like OpenAI insisting that we don’t actually care about the issues but are just being funded to do this.”