OpenAI wants a 'new New Deal' for 2028: an agenda that looks after voters and, along the way, the business

By TechCrunch · November 4, 2026.

Chris Lehane, OpenAI's chief global affairs officer, has published an essay arguing that the 2028 US presidential election could be a defining moment for AI. He calls on politicians to craft policies that address Americans' real concerns, instead of settling for rhetorical gestures in response to the backlash against AI and data centers.

Lehane compares AI's transformative potential to Roosevelt's New Deal, the set of federal programs and reforms from the 1930s. He argues for a "compelling third way" so that people share in AI's benefits. According to TechCrunch, he sets out five priorities ranging from worker protection and education to AI safety and data center construction.

The essay comes after the Democratic victory in the midterm elections, in which the party regained control of Congress amid frustration over the cost of living. According to the article, AI was not a dominant issue, but opposition to data centers was a bipartisan concern in several races.

The proposals, according to TechCrunch, include tax incentives for workers to learn AI, better access to capital for entrepreneurs, and new models of employee ownership and profit-sharing. He also proposes enforceable safety standards, independent evaluations, public oversight, and the ability to restrict AI systems or shut them down with a "kill switch." He further calls for safeguards for minors using AI companions and warns against keeping AI out of classrooms.

Criticism has not been lacking. Kate Brennan, who leads AI policy and corporate power at the AI Now Institute, says the tone marks a shift from OpenAI's agenda last year, which focused on accelerating development, lowering regulatory barriers and expanding infrastructure. Brennan calls the use of the New Deal metaphor "ahistorical at best and harmful at worst": invoking the New Deal means invoking the state's capacity to benefit the public, and this text is signed by a private company advocating policies in its own interest. She also notes that almost everything rests on society already benefiting from AI, something that, she says, we have yet to see.

There are nuances worth remembering. In 2024 OpenAI opposed California's SB 1047 bill, which would have required catastrophic risk assessments and shutdown capability, and last year it tried to limit state regulation. According to the article, it has shifted some of those positions: it backs several state safety laws and says it is open to independent evaluators inside the company. TechCrunch asked OpenAI for more details and received no response before publication.

**Our take.** That one of the most influential companies in the industry acknowledges that economic anxiety is legitimate is good news: the short term will be hard, and denying it only fuels the backlash. It is also right that a blanket ban on AI in classrooms or on data centers carries real costs. But a proposal written by those who stand to benefit from it should be read with one simple question: what obligations does the author take on?

Here lies the weak point. Training incentives and access to capital are voluntary and cheap for companies. What would change the distribution, such as binding wealth-sharing obligations, making companies pay the costs of job displacement, or limits on market concentration, is absent. And the proposal that people retain control over important decisions loses force if AI use becomes mandatory at work, in healthcare or in government.

Overall, the AI debate will enter 2028 through the door of affordability: electricity bills, jobs and who captures the gains. There, OpenAI is trying to set the frame before its critics do. Our view is that the horizon of abundance is plausible, with diseases cured and work chosen, but it will not arrive on its own. It requires institutions that distribute the surplus during the transition, and those institutions must be designed by the public, not by the companies that gain the most. What is useful about the essay is not its content but that it forces the conversation. What we will judge are the concrete, enforceable commitments OpenAI is willing to accept.