Sam Altman and Senator Bernie Sanders do not agree on much. But both have now floated a strikingly similar idea: the pub

Sam Altman and Senator Bernie Sanders do not agree on much. But both have now floated a strikingly similar idea: the public should own a direct financial stake in the artificial intelligence systems trained on their collective work. Altman’s proposal would distribute equity from AI giants like OpenAI to American households, reframing who profits when machines learn from decades of human-created books, code, and art. The concept sounds radical, yet it is best understood not as philanthropy but as an attempt to solve three distinct problems at once—uncompensated data labor, economic anxiety about automation, and the brutal political realities of building trillion-dollar technology companies in a democracy.

The Logic of a National AI Dividend

The first pillar of the argument is straightforward. Large language models like GPT-4 are built by scraping an ocean of human expression—novels, forum posts, music lyrics, software repositories, and photographs—without paying the people who created them. So far, the legal and economic model has treated this as fair use or as an unavoidable byproduct of open publishing. Critics call it extraction. If an oil company drills on public land, it pays royalties. If a streaming platform licenses a song, it pays mechanical rights. AI companies, by contrast, have captured that value upstream and converted it into private equity. Altman’s proposal tacitly admits the injustice: an equity stake functions as a retrospective payment for data labor, acknowledging that the training corpus is not merely ambient noise but the raw material of the product.

The second pillar is economic insurance. Economists remain divided on whether AI will deliver mass unemployment or gradual occupational drift. What is not theoretical is the fear itself. Radiology residents, technical writers, customer service managers, and junior coders are already watching tools automate tasks that once required human judgment. A few hundred dollars in stock will not replace a salary. But the symbolic weight is different. By turning citizens into shareholders, the model attempts to convert ambient dread into aligned interest. You may still lose your job to an LLM, but at least you own a fraction of the LLM. It is the same political logic that underpinned employee stock ownership plans during the rise of the tech industry—broaden the cap table, and you broaden the constituency that roots for the technology’s success rather than its restraint.

Breaking Down the Numbers: A $300 Stake?

To see why this idea is more philosophical than immediately transformative, look at the arithmetic. After its March funding round, OpenAI reached a valuation of roughly $852 billion. A 5 percent equity slice carved out for the public would be worth about $42.6 billion on paper. Spread across approximately 133 million American households, that translates to roughly $320 per household.

Three hundred and twenty dollars will not cover a month of groceries in most zip codes. The figure reveals the core tension. Even a commanding stake in one of the world’s most valuable private companies produces only a modest household payout when the denominator is the entire United States. If the goal is material economic security, the numbers fall short. If the goal is recognition and precedent, the numbers are beside the point.

Then there is the question of form. Directly registering 133 million households as shareholders in a private company is an administrative nightmare. Cap tables are not designed for nation-scale populations. Dividends would trigger 1099s, tax events, and a customer service burden that would make the IRS blush. For these reasons, policy thinkers have looked to the Alaska Permanent Fund as a more practical vessel. Established in the 1970s, that fund pools North Slope oil revenues and pays an annual dividend to every Alaska resident. An AI equivalent would hold the equity in trust, reinvest any proceeds, and distribute cash only when the underlying assets generate sustainable profits.

那个条件——可持续盈利——正是症结所在。OpenAI 目前并非以盈利为优化目标。它的优先级在于增长和资本支出，据报道正向大规模数据中心投入数十亿美元，并推迟任何首次公开募股（IPO），直到公司估值能达到 1 万亿美元。如今持有 OpenAI 股份的主权财富基金持有的是一场投机性赌注，而非一台分红机器。家庭可能需要等待数年才能拿到支票，如果他们最终能拿到的话。

政治博弈与监管红利

在经济逻辑之下，隐藏着冷酷的政治算计。OpenAI 并非在真空状态下运作。民调一致显示，大多数美国人不信任 AI 公司，且由于担心电网压力、耗水量和噪音问题，当地社区对数据中心建设的反对情绪正在上升。Altman 需要获得建设的“社会许可”，而将公众纳入股权结构表（cap table）正是获取这种许可的一种方式。

还有华盛顿方面的问题。该提议提出之际，OpenAI 正寻求与特朗普政府结盟，而该政府已表现出对涉及 Intel 和 Nvidia 等主要芯片制造商的技术中心型经济交易的兴趣。通过将 OpenAI 的成功等同于国家财富，Altman 所做的不仅仅是进步主义式的财富再分配。他正试图让自己被视为国内关键基础设施，而非供应链风险。在华盛顿正竭力在 AI 领域竞争并超越中国发展的时代，任何能将美国民众的财务利益与 OpenAI 的崛起捆绑在一起的模型，都会让监管机构难以通过出口管制、安全审查或反垄断审查来限制该公司。这笔分红不仅仅是金钱，它是一种披着民粹主义外衣的游说策略。

真正的启示

如果这项提议以任何形式取得成功，其留下的遗产将不是那张 320 美元的支票，而是一个先例。这一构想迫使人们进行一场迟到已久的讨论：驱动 AI 的数据究竟应该被视为一种共享的自然资源，还是仅仅作为第一家拥有足够 GPU 进行抓取的公司的免费采石场资源。一个国家级的 AI 基金，无论规模多么微小，都将确立这样一个事实：人类创造的文化具有集体经济价值，这种价值在被喂入训练流水线后依然存在。这一框架比第一笔分红更为重要。它能否在未来五年的游说、会计审计和政治风浪中幸存下来，将决定普通美国人仅仅是 AI 热潮的原材料，还是终于能以股东的身份参与其中。