Andrew Yang said during a debate with AI optimists on the Surrounded program that artificial intelligence is likely to widen job losses and reinforce a winner-take-all market structure dominated by major tech companies. His argument centered on a simple point: companies are using the public’s data and content to train AI models, but the people supplying that material are not being paid.
Yang calls for a data dividend and tax changes
In the debate, Yang said uncompensated public data used by companies is worth more than $300 billion a year, while ordinary people receive none of the profits. He promoted the idea of a Data Dividend, arguing that people who provide content should receive a revenue share when their data or creative work is used to train AI systems.
He also tied that proposal to tax reform. Yang argued that a value-added tax could be used to capture excess profits from large technology companies and help build a stronger economic safety net. In his view, such a system would also address a weakness in traditional corporate income taxes, which he said can be avoided through cross-border transfers and accounting tactics.
AI seen as deepening concentration and job polarization
Yang said AI is driving capital, data and computing resources into the hands of a small group of technology giants, creating a stronger winner-take-all setup. He said AI may lower some technical barriers, but it is also accelerating the decline of mid-tier white-collar roles, naming lawyers, accountants and data analysts as examples.
He argued that large companies are profiting from models trained on free public data while most people do not share in the returns. At the same time, capital is shifting toward AI servers and data centers, and some companies are freezing entry-level hiring as part of workforce reductions. Yang said that trend is pushing up hidden unemployment.
Universal basic income as an economic floor
As labor markets adjust to AI, Yang repeated his support for universal basic income and a data dividend. He said personal data carries major commercial value, and companies that use personal data and creative work to train AI models should pay licensing fees and share the proceeds.
On policy design, he suggested looking to structures such as sovereign wealth funds or weighted indexes of technology companies, with part of the proceeds distributed as a universal basic dividend. He said people who receive those payments would likely spend them in their local communities, helping regional economic circulation.
A personal example of paid AI training use
Yang also cited his own experience. He said an AI company paid him $2,500 in licensing fees through his publisher so the model could learn from the contents of a book he wrote. In his view, if creators were routinely paid similar licensing fees when their work was used for AI training, that would be a fairer economic model.
Why he favors a value-added tax
Yang said conventional corporate income taxes are easier for large technology firms to avoid through multinational profit shifting and accounting arrangements, which weakens the tax base. A value-added tax, he argued, could be imposed directly at each value-creating step in digital services, ad clicks and data transactions, leaving less room for tax avoidance.
He said revenue raised that way could flow into a universal basic income system and turn part of the profits generated by AI into a public good, so the economic gains from technological progress are shared more broadly.
Key points listed in the report
- AI is concentrating capital and data while creating a winner-take-all structure that pressures middle-class white-collar jobs.
- Companies are shifting investment toward data centers and reducing hiring, adding to unemployment pressure.
- Personal data is a core input for AI training and should be covered by licensing and dividend mechanisms.
- Universal basic income is presented as a base layer of economic security, funded in part through data dividends and a value-added tax.
- Traditional corporate income taxes are described as less effective against cross-border tax avoidance than a value-added tax.