观点:别让OpenAI的肥皂剧掩盖了先例

内容来源:https://www.geekwire.com/2026/opinion-dont-let-the-openai-soap-opera-hide-the-precedent/
内容总结:
加州法庭上演科技巨头对决 但AI非营利转营利的核心争议仍需华盛顿裁决
加州奥克兰联邦法院正在上演一场堪比电影情节的法庭大戏:埃隆·马斯克诉山姆·奥特曼,戏剧性的解雇与复职,微软CEO出庭作证谁才是“房间里的大人”。这场庭审吸引着全美目光,但当真正的裁决来临时,最终答案必须来自华盛顿。
硅谷的这出肥皂剧掩盖了一个远比2023年11月那个混乱周末里谁说了什么重要得多的问题——一个关乎每个非营利组织和每个纳税人的问题。作为2014年成立(早于OpenAI一年)且至今仍保持非营利身份的人工智能研究机构艾伦研究所的创始CEO,我希望直截了当地提问:一个依靠税收抵扣捐款资助的慈善机构,能否转化为营利性公司?
如果答案是肯定的,那么美国慈善法就成了一个享有税收优惠的试验场,等待日后哪个项目变得有利可图;而“非营利”这个词,就变成了创始人可以在资产价值足够高时随意定义的概念。如果答案是否定的,那么就应该有人来阻止这种行为。加利福尼亚州总检察长罗伯·邦塔和特拉华州总检察长凯西·詹宁斯选择了不作为。
非营利身份不是晴天可以收起的伞,而是一份永久的承诺。
只要非营利组织能获得部分股权,慈善资产就可以被转移至类似私募股权的结构中——这一先例将塑造未来十年科技领域非营利转营利的模式。该模式在未经司法审查的情况下被接受了,这就是为什么我们现在看到马斯克在法庭上就此提起诉讼。辩方企图把审判焦点变成关于马斯克的一面之词,但OpenAI开创的先例才是真正的问题。
两年来,加州大学洛杉矶分校的法学学者、公民组织Public Citizen、诺贝尔奖得主、加州50家基金会以及OpenAI前员工已经通过信件、简报和请愿书向两位总检察长有力阐述了这一观点。缺少的只是执法行动。
转换的辩护者会告诉你:OpenAI的非营利董事会仍控制着新的营利性公益公司;基金会目前持有约26%的股份,价值约1300亿美元;总检察长们确实争取到了一些让步;而阻止转换会摧毁价值、一事无成。这些论点都有一定道理,前三点甚至部分属实。
但我们别忘了,实验已经做过了,结果就在眼前。
2023年11月解雇山姆·奥特曼,就是对非营利治理是否真实存在的一次实证检验。董事会行使了其最重要的权力——罢免CEO——却在五天内就被持有股权的员工和投资者逼得逆转决定。微软CEO周一作证说,他不得不出面充当“房间里的大人”。董事会刚一试图行使权力,权威便瞬间蒸发。
Public Citizen在2025年9月致两位总检察长的信中明确表示:这个非营利组织已经“沦为了营利实体的橡皮图章”。
这就是萨克拉门托(加州首府)和威尔明顿(特拉华州首府)本应面对但未予追究的检验案例。10月28日,邦塔发布和解声明,基于“非营利控制权得到保留”的表述批准了资本重组。同一天,特拉华州的詹宁斯几乎逐字逐句地复制了加州的声明,发布了“无异议声明”。当和解声明告诉你非营利控制权得到保留时,请问清楚是哪个版本——纸面上的版本,还是接受检验时真正站得住的版本。
更深层的问题是,此案为何最终选择了和解。两地总检察长办公室都拥有法律依据、资源和管辖权。他们之所以没有动用这些权力,是出于政治考量——两位州总检察长,各自手下只有一个慈善机构事务局,却要权衡是否要与史上最有价值的私人公司、其首席投资者以及一个将AI主导地位明确列为产业政策的美国政府对抗。
法官伊冯·冈萨雷斯·罗杰斯仍有可能下令撤销资本重组、罢免奥特曼和布罗克曼、将罚款返还给慈善机构。其中任何一项裁决都具有重要意义。但此案围绕的是马斯克的具体诉求和具体法律地位,即便做出完全有利于他的裁决,也无法约束下一位面临类似转换案件的州总检察长。这个先例问题将比判决持续更久。
解决方案在于联邦层面。州总检察长已经表态;前进道路通往华盛顿,国税局、国会两院的税务委员会以及联邦反垄断机构都有发挥作用的可能。一个联邦案例可以确立一个原则:下一个考虑进行此类操作的AI实验室将面临后果。
这个先例的影响远超人工智能领域。它关乎每一个非营利组织,也关乎每一位因为相信规则有其意义而选择走更难道路的创始人。我们需要有勇气执行法律、保护非营利体系的公职人员。萨克拉门托和威尔明顿失败了。华盛顿还没有尝试过。
披露: 本文作者是《纽约时报》诉OpenAI及微软版权案中原告方的聘用专家。所表达观点仅代表其个人立场。
中文翻译:
我们正在奥克兰观看一场电影级别的法庭大戏:埃隆·马斯克对阵萨姆·奥尔特曼,一场戏剧性的解雇与复职,微软首席执行官出庭作证,讨论谁是“房间里的大人”。这场闹剧让我们的目光紧盯着法庭。然而,最终的清算,终究要来自华盛顿特区。
硅谷的这出肥皂剧掩盖了一个远比“2023年11月那个混乱周末里谁说了什么”更重要的问题。这个问题关系到每一家非营利组织和每一位纳税人。作为艾伦人工智能研究所——一个成立于2014年(比OpenAI早一年)、至今仍是非营利机构的研究实验室——的创始CEO,我想直截了当地问:一个由可抵税捐赠资助的慈善机构,能被转化为一家营利性公司吗?
如果你的答案是肯定的,那么美国慈善法就成了一个享有税收优惠的“跳板”,任何后来证明有利可图的风险投资都可以从这里起步;而“非营利”这个词,在资产变得足够值钱后,就成了创始人随心定义的东西。如果你的答案是否定的,那么就应该有人来阻止这一切。加州总检察长罗布·邦塔和特拉华州总检察长凯西·詹宁斯选择了不作为。
非营利身份不是一出太阳就可以收起的雨伞。这是一份持久的承诺。
那个先例——只要有一部分股权回流到非营利机构,慈善资产就可以被转移到私募股权式的结构中——将塑造未来十年科技领域非营利转营利的风向。它未经司法审查就被接受了,这就是为什么我们现在看到埃隆·马斯克在法庭上争论这个问题。辩护方正试图把审判变成对他个人的指控,但OpenAI所立的先例才是真正的问题。
两年来,加州大学洛杉矶分校的法学学者、公共市民组织、诺贝尔奖得主、加州五十家基金会以及OpenAI前员工,通过信件、法律摘要和请愿书,向两位总检察长强烈提出了这一论点。缺失的是执法。
转型的辩护者会告诉你:OpenAI的非营利董事会仍然控制着新的营利性公益公司;该基金会目前持有26%的股份,价值约1300亿美元;两位总检察长已争取到了实质性的让步;而阻止转型的替代方案则会摧毁价值,一无所获。这些论点每个都有一定道理,前三个甚至部分属实。
但且慢。我们已经做过实验了,结果已经出炉。
2023年11月解雇萨姆·奥尔特曼,正是对非营利治理是否真实存在的一次实证检验。董事会行使了它最重要的权力——罢免CEO——却在五天内被持有股权的员工和投资者逼得逆转决定。微软首席执行官周一作证时表示,他不得不作为“房间里的大人”介入。董事会的权威在它试图行使的那一刻就蒸发了。
公共市民组织在2025年9月写给两位总检察长的信中直言不讳:这个非营利机构已经“沦为营利实体的橡皮图章”。
这就是萨克拉门托和威尔明顿面前的那个测试案例,也是他们没有追究的案例。10月28日,邦塔发布了一份和解声明,基于非营利控制权得以保留的陈述,批准了资本重组。同日,特拉华州的凯西·詹宁斯发布了一份“不反对声明”,几乎逐字逐句地效仿了加州表述。所以,当和解方案告诉你非营利控制权得以保留时,请问一问是哪个版本——纸面上的版本,还是经得起实际考验的版本?
更深层的问题是,这个案子为什么会被和解。两个总检察长办公室都具备诉讼资格、资源和管辖权。他们不用的原因是政治性的——两位州总检察长,各自只有一个慈善事务局,却要权衡是否对抗历史上最有价值的私营公司、它的主要投资者,以及一个将人工智能主导地位明确列为产业政策的白宫。
法官伊冯·冈萨雷斯·罗杰斯仍然可以下令撤销资本重组、罢免奥尔特曼和布罗克曼、将损害赔偿金返还给慈善机构。任何一项裁决都意义重大。但此案的关键在于马斯克的具体诉求和诉讼资格,即使做出对他有利的全面裁决,也无法约束下一位考虑下一个转型案例的州总检察长。这个先例问题将比判决本身更持久。
解决方案在于联邦层面。州总检察长已经表态了;前进的道路通向华盛顿特区。在那里,国税局、两院的税务制定委员会以及联邦反垄断机构,都有各自可能发挥的作用。一个联邦案例可以确立:下一个考虑此类操作的人工智能实验室将面临后果。
这个先例的影响远超人工智能领域。它关系到每一个非营利组织,也关系到每一位因为相信规则有意义而选择了更艰难道路的创始人。我们需要有勇气的公职人员来执行那些保护非营利体系的法律。萨克拉门托和威尔明顿失败了。华盛顿还没有尝试过。
披露声明:作者在《纽约时报》诉OpenAI和微软版权诉讼案中担任原告方聘请的专家。文中观点仅代表其个人。
英文来源:
We are watching a courtroom drama in Oakland made for motion pictures. Elon Musk versus Sam Altman, a dramatic firing and reinstatement, the CEO of Microsoft testifying about who was “the adult in the room.” The spectacle is keeping our eyes on the courtroom. Yet the reckoning, when it comes, will have to come from Washington, D.C.
The Silicon Valley soap opera is obscuring a question that matters far more than who said what during a chaotic weekend in November 2023. It is a question that should concern every nonprofit, and every taxpayer. As the founding CEO of the Allen Institute for AI — a nonprofit research lab launched in 2014, a year before OpenAI, and still a nonprofit today — I want to ask it plainly: can a charity, funded by tax-deductible donations, be converted into a corporation?
If your answer is yes, then American charity law is a tax-advantaged staging ground for whichever venture later proves lucrative, and the word “nonprofit” means whatever the founders decide it means once the asset gets valuable enough. If your answer is no, then someone should stop this. California Attorney General Rob Bonta and Delaware Attorney General Kathy Jennings chose not to.
Nonprofit status is not an umbrella you fold once the sun comes out. It’s a lasting commitment.
That precedent — that charitable assets can be migrated into private-equity-style structures as long as some equity stake flows back to the nonprofit — will shape the next decade of nonprofit-to-for-profit conversions in technology. It was accepted without judicial review, and that is why we are now watching Elon Musk litigate the question. The defense is making the trial about him. The precedent OpenAI is setting is the real issue.
This argument has been made forcefully by UCLA legal scholars, Public Citizen, Nobel laureates, fifty California foundations, and former OpenAI employees for two years, in letters, briefs, and petitions to both AGs. What’s missing is enforcement.
The defenders of the conversion will tell you that OpenAI’s nonprofit board still controls the new for-profit public benefit corporation, that the foundation now holds a 26 percent stake worth roughly $130 billion, that the AGs extracted real concessions, and that the alternative — blocking the conversion — would have destroyed value and accomplished nothing. Every one of these arguments has some merit. The first three are even partially true.
But hold on. We already ran the experiment. The result is in.
The November 2023 firing of Sam Altman was the empirical test of whether nonprofit governance was real. The board exercised its single most important power — removing the CEO — and was forced into reversal within five days by employees and investors who held the equity. Microsoft’s CEO testified on Monday that he had to step in as “the adult in the room.” The board’s authority evaporated the moment it tried to use it.
Public Citizen put it plainly in its September 2025 letter to the two AGs: the nonprofit has become “little more than a rubber stamp of the for-profit.”
That is the test case Sacramento and Wilmington had in front of them, and the one they did not pursue. On October 28, Bonta issued a settlement statement blessing the recapitalization on the representation that nonprofit control is preserved. Kathy Jennings in Delaware released a Statement of No Objection the same day, tracking California almost word for word. So when the settlement tells you nonprofit control is preserved, ask which version — the version on paper, or the version that actually held when tested.
The deeper question is why the case settled at all. Both offices had standing, resources, and jurisdiction. The reasons they didn’t use them are political — two state AGs, each with a single charity bureau, weighing a fight against the most valuable private company in history, its lead investor, and a White House that has made AI dominance an explicit industrial policy.
Judge Yvonne Gonzalez Rogers could still order an unwinding of the recapitalization, the removal of Altman and Brockman, damages funneled back to the charity. Any of those would matter. But the case turns on Musk’s particular claims and Musk’s particular standing, and even a sweeping ruling in his favor would not bind the next state attorney general weighing the next conversion. The precedent problem outlives the verdict.
The solution is a federal case. The state AGs have spoken; the path forward runs through Washington, where the IRS, the tax-writing committees in both chambers, and the federal antitrust agencies each have a potential role to play. A federal case can establish that the next AI lab considering this maneuver will face consequences.
The precedent matters far beyond AI. It matters for every nonprofit, and for every founder who took the harder road because they believed the rules meant something. We need public officials with the courage to enforce the laws that protect the nonprofit system. Sacramento and Wilmington failed. Washington has not yet tried.
Disclosure: The author serves as a retained expert for the plaintiffs in The New York Times v. OpenAI and Microsoft copyright litigation. The opinions expressed are his alone.