On August 19th at 7:07PM, Tyler Johnston received a message from his roommate. A man, the text said, was knocking on their door with legal documents to serve.<\/p>\n<\/div>\n
Johnston is the founder and sole employee of The Midas Project: a nonprofit that monitors the practices of \u201cleading AI companies to ensure transparency, privacy, and ethical standards are maintained.\u201d The Midas Project is behind The OpenAI Files<\/a>, a 50-page report about OpenAI\u2019s evolution from under-the-radar nonprofit to moneymaking household name. It organized an open letter to OpenAI<\/a> asking for transparency about its transition to a for-profit company, garnering more than 10,000 signatures. Now, apparently, OpenAI was striking back.<\/p>\n<\/div>\n Johnston was traveling in California at the time, but he quickly learned that OpenAI had hired a process server from an Oklahoma-based firm called Smoking Gun Investigations, LLC, whose website featured a smoking pistol and the tagline, \u201cA bitter truth is better than the sweetest lies.\u201d Two 15-page subpoenas ended up being delivered by email: one to The Midas Project, the other to Johnston personally. The documents suggested The Midas Project was a catspaw of OpenAI\u2019s longtime rival Elon Musk \u2014 dragging the nonprofit into a bitter lawsuit between the pair.<\/p>\n<\/div>\n Johnston wasn\u2019t surprised that OpenAI\u2019s lawyers had come calling, he told The Verge<\/em> in an interview. But the \u201cegregious\u201d breadth of the subpoenas surprised him. OpenAI wasn\u2019t just demanding to know whether his tiny nonprofit had gotten funding from Musk. It wanted to know all of its funding sources, right down to when and how much donors contributed. And it wanted the nonprofit to turn over any documents and communications related to OpenAI\u2019s governance and structure and any potential changes to it.<\/p>\n<\/div>\n \u201cI didn\u2019t mind if they asked if we were funded by Musk,\u201d Johnston said. \u201cBut they could\u2019ve looked up our form 990 and seen we brought in less than $75,000 last year. If Musk were funding us, he\u2019d have to be very thrifty about it.\u201d<\/p>\n<\/div>\n Johnston\u2019s experience isn\u2019t unique. In recent months, OpenAI has subpoenaed a wide range of nonprofits that have been critical of the company\u2019s controversial for-profit restructuring. Ostensibly, the subpoenas are supposed to help OpenAI build its defense against Musk, who sued to stop the company\u2019s transition. But in practice, recipients and legal experts say, they seem more like a campaign of intimidation with very real costs. Online controversy has roiled OpenAI, with current and former employees publicly criticizing the company\u2019s legal tactics. And at a time when AI companies are garnering unprecedented money and power, the subpoenas call attention to OpenAI\u2019s ongoing departure from its nonprofit roots.<\/p>\n<\/div>\n OpenAI did not provide a comment by publication time. A company executive has previously said on X that there was \u201ca lot more to the story\u201d than the nonprofits claim.<\/p>\n<\/div>\n So far, at least seven nonprofits have revealed that they received subpoenas, including the San Francisco Foundation, Encode, Ek\u014d, the Future of Life Institute, Legal Advocates for Safe Science and Technology, and the Coalition for AI Nonprofit Integrity. Many of the subpoenas seem to request not only answers to whether the nonprofits are involved with or funded by Musk in any way, but also every entity that has ever financially supported the nonprofits, as well as every one of the nonprofits\u2019 documents and communications related to OpenAI\u2019s own restructuring. The requests were wide-ranging enough to bury some of the nonprofits with paperwork and legal fees.<\/p>\n<\/div>\n The outcry has been far-reaching, with even OpenAI\u2019s own mission alignment team lead, Joshua Achiam, posting on X<\/a>, \u201cAt what is possibly a risk to my whole career I will say: this doesn\u2019t seem great.\u201d<\/p>\n<\/div>\n Achiam added, noting that he was speaking in a personal capacity, \u201cWe can\u2019t be doing things that make us into a frightening power instead of a virtuous one. We have a duty to and a mission for all of humanity \u2026 I would not be at OpenAI if we didn\u2019t have an extremely sincere commitment to good. But there are things that can go wrong with power and sometimes people on the inside have to be willing to point it out loudly.\u201d<\/p>\n<\/div>\n The subpoenas weren\u2019t just an inconvenience. The OpenAI Files\u2019 source material included confidential conversations with former OpenAI employees and other sources close to the company, and communications around the open letter included private information about its signatories. Providing what OpenAI requested would put all that in its hands.<\/p>\n<\/div>\n James Grimmelmann, a professor at Cornell Law School and Cornell Tech, told The Verge<\/em> that it\u2019s \u201creally hard\u201d to see how determining whether Musk funded nonprofits to attack OpenAI would be relevant to the lawsuit, especially given how speculative the links appear.<\/p>\n<\/div>\n \u201cThese requests are going to be chilling in terms of turning over that much information to this powerful a company,\u201d Grimmelmann said. \u201cThese require really extensive searches through these organizations\u2019 records and very detailed responses that are going to be quite expensive. They\u2019re going to rack up extensive legal fees responding. Or if they move to quash the subpoena, those are going to be equally expensive legal fees fighting with OpenAI.\u201d<\/p>\n<\/div>\n If OpenAI wanted to demonstrate good faith, Grimmelmann said, it could have offered to pay the nonprofits\u2019 legal fees. As it stands, he compared the subpoenas to Musk\u2019s own ruinous litigation campaigns against nonprofits critical of X.<\/p>\n<\/div>\n Johnston found a lawyer and chatted with other nonprofits facing the same obstacles from OpenAI. He learned quickly that legal professionals considered the requests \u201cunreasonable\u201d and that he didn\u2019t need to produce all the documents. He also learned that even the court seemed skeptical of OpenAI\u2019s demands. In August, a judge who had allowed the company to pursue discovery over Musk\u2019s acquisition efforts and past plans for OpenAI said the court was reconsidering the decision, \u201chaving seen the scope of the discovery and potential discovery that OpenAI is attempting to drive through this opening.\u201c<\/p>\n<\/div>\n But although Johnston was able to, for now, avoid providing the full extent of the documents requested, being targeted by OpenAI has done tangible damage. He tried to get legal insurance after the incident, only to be universally turned down by insurers, in at least one case explicitly because they were concerned about the OpenAI-Musk dispute. \u201cIt kind of made us uninsurable, and that\u2019s another way of constraining speech,\u201d he said.<\/p>\n<\/div>\n Nathan Calvin, another recipient of a subpoena, is general counsel at Encode \u2014 a policy nonprofit that worked on California\u2019s landmark AI safety law, SB 53. He was served when a sheriff\u2019s deputy knocked on his door during dinner, and was so surprised that he recalls fist-bumping the sheriff\u2019s deputy and telling him to have a good night. As with Johnston, one subpoena went to Calvin personally, while another was served to Encode.<\/p>\n<\/div>\n Encode was able to retain pro bono legal counsel, which Calvin said saved the three-person nonprofit thousands, or tens of thousands, of dollars. But Calvin was troubled by an odd detail in the subpoena: a request for all documents and communications regarding SB 53 or its \u201cpotential impact on OpenAI.\u201d OpenAI had publicly fought<\/a> the legislation. What, Calvin wondered, might it gain from learning how Encode had lobbied for the law \u2014 and seeing who it had spoken with to do so?<\/p>\n<\/div>\n After the nonprofit went public with the subpoena, OpenAI CSO Jason Kwon posted a statement<\/a>, saying there was \u201cquite a lot more to the story\u201d than Calvin\u2019s viral X post<\/a>. \u201cAs everyone knows, we are actively defending against Elon in a lawsuit where he is trying to damage OpenAI for his own financial benefit,\u201d Kwon wrote. He listed out points on why OpenAI had subpoenaed Encode specifically, focusing largely on the fact that Encode filed an amicus brief<\/a> opposing OpenAI\u2019s restructuring in the Musk v. Altman<\/em> case. He also wrote that OpenAI \u201cdid not oppose SB53; we provided comments for harmonization with other standards.\u201d<\/p>\n<\/div>\n In a follow-up post<\/a>, Kwon said the company had subpoenaed the other nonprofits because they had also seemed to oppose its restructuring and could in theory be backed by Musk.<\/p>\n<\/div>\n When asked about Kwon\u2019s post and reasoning, Cornell\u2019s Grimmelmann said, \u201cThat might be an appropriate approach when you are doing large-scale corporate litigation between behemoths like X and Openai, but it\u2019s really oppressive to target nonprofit organizations that way.\u201d<\/p>\n<\/div>\n The SB 53 inquiry wasn\u2019t the only policy-focused request. OpenAI\u2019s requests for Legal Advocates for Safe Science and Technology reportedly<\/a> included<\/a> turning over all documents and communications concerning SB 1047, an earlier safety framework that Newsom vetoed in 2024, or AB 501<\/a>, a bill that may have barred OpenAI\u2019s for-profit conversion.<\/p>\n<\/div>\n Sacha Haworth, executive director of the Tech Oversight Project, which co-released The OpenAI Files with The Midas Project, told The Verge<\/em> that OpenAI is using \u201clawfare\u201d to its advantage, adding that the company is making \u201cparanoid accusations about the motivations and funding of these advocacy organizations that for the most part have nothing to do with OpenAI.\u201d (Tech Oversight Project, so far, has not reported being served by OpenAI.)<\/p>\n<\/div>\n \u201cOpenAI had a chance, in the last couple of years, to differentiate themselves from the rest of the Big Tech pack in the way that they operate politically, and it\u2019s clear that they\u2019re following in the footsteps of the Metas and the Amazons,\u201d Haworth said. \u201cThat\u2019s the playbook that they\u2019re following, and they use lawyers and political donations, super PACs, lobbyists \u2014 they use it all. Because they can. And sometimes it works.\u201d<\/p>\n<\/div>\n The Future of Life Institute confirmed to The Verge<\/em> that it also received a subpoena in the Musk v. Altman<\/em> legal battle on October 2nd, but its president, Max Tegmark, personally received one in August. \u201cLike those served to Encode and others, the subpoenas are quite broad,\u201d Anthony Aguirre, cofounder and executive director of the institute, told The Verge<\/em> in a statement. \u201cWe remain laser-focused on our mission of steering AI away from large-scale risks and toward beneficial outcomes for humanity.\u201d<\/p>\n<\/div>\n Judith Bell, the San Francisco Foundation\u2019s chief impact officer, told The Verge<\/em> that the subpoena it received was also \u201cquite over-broad\u201d with \u201clots of over-reach.\u201d It had a few specific questions about the SFF\u2019s relationship with Musk, Mark Zuckerberg, and Meta, but also requested \u201cto receive all of our information about anything connected to OpenAI\u2019s structure, governance, et cetera,\u201d Bell said. It was \u201cclearly a fishing expedition of them trying to figure out what we\u2019re doing strategically\u201d regarding OpenAI\u2019s for-profit conversion, she added.<\/p>\n<\/div>\n Bell said the SFF had actually met with representatives of OpenAI months earlier, seeking details about its for-profit conversion. (According to her, OpenAI\u2019s representatives did not answer the questions.) As part of that meeting, they\u2019d told OpenAI they had no relationship, financial or otherwise, with Musk.<\/p>\n<\/div>\n The subpoenas aren\u2019t the big story here, says Bell \u2014 OpenAI\u2019s restructuring is. Among other goals, SFF is asking the attorney general to protect the OpenAI nonprofit\u2019s charitable assets with independent governance and oversight as the company restructures. Those assets \u201care owned by the public,\u201d Bell says. SFF has also questioned how much equity the nonprofit will receive in the new structure and how a for-profit OpenAI will be governed. These tactics, by comparison, are \u201ca bit of a distraction technique.\u201d<\/p>\n<\/div>\n \u201c[You] end up with a story about the subpoenas instead of a story of what\u2019s really at stake here,\u201d she said.<\/p>\n<\/div>\n\n