OpenAI Funds a Study of How AI Could Shift Tax Revenue

OpenAI is funding a study of how artificial intelligence could be taxed, but the public commitment is narrower than a call for a new AI levy. The Tax Foundation is being asked to assess tax options and examine whether AI adoption changes the mix of public revenue from labor income, corporate profits and capital gains; neither a preferred policy nor the grant amount has been disclosed.

The recipient is the Tax Foundation, described as a U.S.-based global tax think tank. It will use OpenAI’s grant for research and a paper evaluating AI tax options against five stated criteria: neutrality, simplicity, transparency, stability and economic efficiency.

The study combines two related but different questions. One concerns the design of taxes applied to AI. The other concerns where income and revenue may arise if AI adoption changes the balance among wages, corporate profits, capital gains and other public revenues.

That distinction leaves room for more than one outcome. A paper assessing AI tax options could examine an AI-specific charge. It could also assess whether changes in income and profits are better addressed through existing tax structures. The announced research does not identify the tax instruments to be compared or say which approach the Tax Foundation expects to favor.

The five criteria provide the study’s stated yardstick, but not its conclusion. A system can be simple to administer yet produce different revenue effects than another system; the public description does not explain how the paper will weigh such tradeoffs. It also does not describe the research methods that will be used to measure AI’s effects on income or public revenue.

$1 millionCash grants

OpenAI awarded $1 million in cash across 14 independent projects studying AI’s economic and social effects.

Up to $1 millionAPI credits

OpenAI also provided up to $1 million in API credits across the selected projects.

Fewer than 3.5%Selection rate

More than 400 people and organizations submitted proposals, and fewer than 3.5% of applicants were selected.

The tax paper is one of 14 independent projects in OpenAI’s wider program on AI’s economic and social effects. OpenAI awarded $1 million in cash across the group and offered up to another $1 million in API credits, but did not disclose how either pool was divided among recipients. That leaves the size of the Tax Foundation award unknown even within the larger program.

OpenAI initially offered focused grants of up to $100,000 through its call for proposals. The selected projects are scheduled to run for six months and report results in 2027. The program therefore establishes a period of funded research, not completed findings on the tax consequences of AI adoption.

OpenAI has described the program as following its Industrial Policy for the Intelligence Age and as funding policy development outside the company. Nine projects focus on economic opportunity, while five address societal resilience and safety. The work spans projects focused on the United States, the European Union, Brazil, Singapore and South Korea.

The tax research sits beside work on workforce disruption, benefits, the distribution of AI productivity gains and power-system expansion for data-center demand. Those are separate assignments, but together they frame AI as a policy issue affecting both the sources of public revenue and the public systems that may respond to economic change.

Other projects in the portfolio address

• Workforce scenarios connecting low-, moderate- and high-disruption labor-market outcomes to possible policy responses.
• A person-based benefits system covering retirement, health, leave, education, training and disability.
• How AI productivity gains are distributed among workers, companies, industries and regions, and how U.S. states could expand power generation and transmission for data-center demand.
• Safety work on recursively self-improving AI, AI-and-biotechnology risks and frontier-AI safety programs for national-security professionals.

The disclosed facts establish that OpenAI selected and funded the Tax Foundation project from the proposal pool. They do not describe editorial controls, review rights, publication commitments, conflicts policies or procedures for handling disagreements between the think tank and its funder. That is a limit of the public accounts, not evidence that such procedures do not exist.

The reporting also provides no recipient-specific amount, methodology or stated publication process for the tax paper. The useful test will come with the research itself: whether it defines the tax options, shows how it evaluates the stated criteria and explains its account of potential revenue shifts. For now, OpenAI has funded an examination of choices that could shape AI policy, not announced the choice it wants governments to make.