The "Industrial Policy for the Intelligence Age" document was put together by Altman, Chris Lehane (head of global affairs), and Josh Achiam, OpenAI's Chief Futurist. That last role appeared in February, when OpenAI disbanded its second safety team and renamed its leader a futurist. Now he writes policy papers.
Altman, in an interview with Axios, compares the scale to Roosevelt's New Deal. The document is addressed to the US government, but OpenAI says the conversation should go global. In May they're opening a workshop in Washington. Here's what they propose to discuss:
Public wealth fund. AI companies contribute to a national fund, which invests in the economy, and every citizen receives dividends from growth — even if they don't own stock.
Shift to capital taxation. If AI replaces jobs, payroll taxes stop funding pensions and healthcare. OpenAI proposes taxing corporate profits and capital income rather than labor. Plus introducing a separate tax on automated work. The EU has been discussing this idea since 2017 (a bill by MEP Mady Delvaux); in January 2026, Brookings published a framework document on AI taxes. OpenAI takes the same arguments and packages them as its own agenda.
32-hour workweek at full pay. Pilot programs with unions. AI boosts productivity, and the gains are returned to workers as free time. OpenAI calls this an "efficiency dividend."
Access to AI as a basic right. Schools, libraries, small businesses, low-income areas — free or cheap access to foundation models, similar to electricity and the internet.
Portable benefits. Healthcare, pension, and training tied to the person, not the employer. Change jobs — everything comes with you.
Automatic social support. The government tracks AI unemployment metrics in real time. When indicators cross a threshold — payments, wage insurance, and retraining vouchers kick in automatically. When they drop — they turn off.
AI containment protocols. An AI system is autonomous, self-replicating, and can't be recalled. Pre-agreed protocols, coordination with governments, and international incident data sharing would help.
Mandatory audits of frontier models. Independent reviews before and after release, but only for the most powerful systems. Leave the rest of the market alone.
*P.S. I wouldn't forget that OpenAI is heading toward an IPO by end of 2026 with a projected $14B loss. They then shut down Sora and lost the Disney contract. And yet they just closed a $122B round. So Altman has every reason to hype up the importance of the technology to raise money at high valuations.*
Breakdown from Axios, document on the OpenAI website.
🤔 A new social contract
🤡 A new round of hype