OpenAI Seeks to Shape 2028 AI Rules and Elections

Across 22 major federal agencies, nearly 378,000 employees left during 2025, while about 127,000 were hired. By January 2026, the agencies had more than 250,000 fewer workers than in December 2024. Federal agencies expanded their use of AI, too, but no consistent public method showed whether services improved. That’s a leaner state taking shape: fewer public employees, more automated systems and a fight over who sets the rules.

Who Gets to Set the Rules

Government Accountability Office figures cited in Federal News Network commentary by Gene Lockwood-Shabat showed AI use cases across 11 federal agencies nearly doubled, from 571 in 2023 to 1,110 in 2024. Reported generative AI use cases climbed from 32 to 282. Yet the commentary said the government lacked a consistent public method to judge whether those services improved.

The Internal Revenue Service said its Research, Applied Analytics and Statistics Division lost 63 employees, most of them directly involved in designing, developing or overseeing AI projects. IRS officials also said the agency hadn’t adequately identified the AI skills it needed. In June 2026, the Department of Veterans Affairs inspector general found that staff used general-purpose generative AI chat tools for clinical work. The VA couldn’t measure how widely staff used the tools or readily identify AI-related patient-safety events. The VA concurred with the inspector general’s recommendations.

OpenAI chief global affairs officer Chris Lehane has stepped into this widening gap between adoption and accountability, pressing for policies ahead of the 2028 presidential election. In an essay, he compared AI’s potential economic effects to President Franklin D. Roosevelt’s New Deal and proposed a “compelling third path” for sharing AI’s benefits. His priorities included worker protections, education, AI safety and data-center development.

Kate Brennan, who leads AI and corporate power policy at the AI Now Institute, said Lehane’s position differed from OpenAI’s policy agenda the previous year, which emphasized accelerating AI development, reducing regulatory barriers and expanding infrastructure. “Invoking a New Deal metaphor is to invoke state capacity to actually benefit the public,” Brennan said. She called the essay “written by a private company that is trying to advocate for policies for its own benefit. It is ahistorical at best and harmful at worst.”

Promises, Without a Binding Share

Lehane proposed incentives for workers to learn AI skills, easier access to capital for entrepreneurs, and new approaches to employee ownership and profit-sharing. Brennan said the proposals didn’t require companies to share AI-generated wealth or pay costs for workers displaced by AI. “I was struck by how much of this policy is for a future...that is already predicated on mass AI adoption and AI benefitting society, which we have yet to see,” she said.

Brennan suggested a revised corporate tax rate, incentives to hire people rather than automate jobs, and measures to prevent companies from building excessive market power. The dispute is plain: will firms driving adoption face binding limits, or will public policy rely on incentives and promises?

Lehane acknowledged concerns about data centers but emphasized construction jobs and local tax revenue. He warned politicians that supporting a moratorium could jeopardize U.S. leadership over China, comparing that position to the 2002 vote to authorize the Iraq war. He also called for age-appropriate protections and safeguards for AI companions used by minors. Keeping AI out of classrooms, he argued, could increase inequality between wealthy students and “children in rural communities.” Critics said that claim assumed AI improves learning, which remains unproven.

Lehane called for enforceable AI safety standards, independent evaluations, public oversight, and the ability to restrict systems or shut them down with a kill switch. Brennan called those positions “disingenuous,” citing OpenAI’s history of opposing legislation intended to impose such measures. OpenAI opposed California’s SB 1047 in 2024; the bill would have required catastrophic-risk evaluations and shutdown capabilities. The company later sought to limit state regulation, then backed several state safety laws and said it was open to independent evaluators embedded in the company. TechCrunch said OpenAI didn’t respond to a request for more details by publication.

Money in the Electoral Machinery

Political groups for and against the AI industry spent millions trying to influence the midterms and future AI regulation. AI-focused PACs American Mission and Building the Future spent a combined $4.8 million backing South Carolina Senate candidate Darlene Graham, who lost. American Mission is the Republican spending arm of Leading the Future, backed by OpenAI President Greg Brockman and venture-capital firm Andreessen Horowitz. It also spent more than $500,000 each supporting losing Republican House candidates Aaron Flint in Montana and Laurie Buckhout in North Carolina. Leading the Future’s Democratic spending arm, Think Big, backed Florida Rep. Jared Moskowitz, who won reelection.

The spending had mixed results elsewhere, too. Public First, backed by Anthropic, supported several Republicans who lost through an affiliated PAC, and had more success backing Democrats. It contributed $3 million to a PAC supporting California state Sen. Scott Wiener, who was leading when the article was published. Guardrails Alliance, which seeks to make accepting AI company money politically costly, spent more than $1 million each opposing California Rep. Jimmy Gomez and Montana’s Flint; both lost. It spent $300,000 supporting Angela Gonzales-Torres, who defeated Gomez. Elon Musk, who reportedly spent half a billion dollars supporting Republicans, also saw many of his candidates lose.

Leading the Future said primaries were a more effective use of its money than general elections. It spent $27.38 million in primaries, including $8 million opposing New York Assembly member Alex Bores, who lost his congressional bid in June. Charlie Blaettler, political director of Guardrails Alliance, said of the group’s comparatively small general-election spending: “They don’t want their money to become an issue in these elections.”

Blaettler pointed to the Michigan Senate primary, where AIPAC spent close to $30 million supporting Democratic candidate Haley Stevens against Abdul El-Sayed. Stevens conceded the primary to El-Sayed, who went on to win the Senate seat. “AIPAC became the story in that race,” Blaettler said. “We know people hate AI billionaires. They hate the people running these companies. They’re very distrustful and disgusted by the political spending and lobbying that the tech industry does. If they spend in a competitive race, they would just become an enormous liability for their candidates.” He said, “Our goal is to make AI money politically toxic.”

The PACs disagree over whether Congress should impose a national framework overriding state laws or leave states free to set safeguards. The fight remains in electoral and legislative channels. Leading the Future favors a single national framework and plans to spend in the 2028 elections; Guardrails Alliance says it will keep opposing candidates backed by the AI industry. Federal News Network’s commentary proposed a public scorecard instead: agencies would report spending, staffing and skills, performance against service targets, and quality-control data, including AI-related errors and administrative rework. Lockwood-Shabat said Congress and the White House’s Office of Management and Budget could build on existing agency plans and budget justifications, rather than create a separate reporting bureaucracy.

“AI may well allow a leaner federal government to accomplish more for taxpayers,” Lockwood-Shabat said. “But faster isn’t inherently better, and fewer employees isn’t a guaranteed win for efficiency. If Washington’s going to expand AI while shrinking the civil service, it must be brave enough to score the results.”