AI Industry Fractures as Altman Lobbies Against Pre-Deployment Reviews
OpenAI's CEO pushes lawmakers to reject model approval requirements while rivals pour millions into competing regulatory visions, exposing deep divisions over governance as US-China competition intensifies.
Sam Altman met with US lawmakers on 3 June 2026 to advocate against proposals requiring government approval before AI developers release new models, positioning OpenAI squarely against mandatory pre-deployment reviews as the industry splits into warring regulatory camps.
The lobbying effort, which included a meeting with House Speaker Mike Johnson, marks a sharp escalation in the battle over AI Governance. Altman called pre-approval requirements “disastrous” for the industry during Senate testimony in May 2025, warning that “a patchwork of regulatory overlays will cause friction in the ability to build and extend what we’re doing.” Now, thirteen months later, that rhetorical opposition has become active Congressional engagement at the highest levels.
The timing is deliberate. Just one day before Altman’s Capitol Hill meetings, President Trump signed an executive order establishing a voluntary 30-day review window for frontier AI models with certain cyber capabilities. Altman is now asking Congress to increase funding for AI testing at the Commerce Department, which already works with OpenAI and Anthropic on voluntary evaluations—but he’s drawing a hard line against making those reviews mandatory.
The Coalition War
Behind Altman’s opposition lies an industry fractured by competing visions of regulatory control. Anthropic quadrupled its federal lobbying spending to more than $3.1 million in 2025 and donated $20 million to Public First Action, a new 501(c)(4) calling for tougher AI Regulation, per OpenSecrets. Public First Action raised nearly $50 million in February 2026.
On the opposite flank, Leading the Future PAC—backed by OpenAI’s Greg Brockman and other tech leaders—announced it raised $125 million to influence the midterm elections. The financial firepower dwarfs Anthropic’s efforts and signals OpenAI’s intention to shape Congressional composition, not just policy.
| Company/Coalition | Regulatory Position | Financial Commitment |
|---|---|---|
| Anthropic | Mandatory pre-deployment safety reviews | $3.1M lobbying (2025), $20M to Public First Action |
| OpenAI/Leading the Future | Voluntary testing, oppose pre-approval | $125M PAC for midterm elections |
| xAI, Google DeepMind, Microsoft | Voluntary CAISI testing agreements | Testing partnerships announced May 2026 |
The split extends to government contracting. In February 2026, Anthropic walked away from a $200 million Pentagon contract rather than loosen safety guardrails on its Claude model, a decision described by safety researchers as proof the company is “willing to put a lucrative contract on the line to signal their commitment to building safer AI models.”
Altman has acknowledged the competitive implications. In remarks captured by Lawfare, he stated: “There will be at least one other actor, which I assume will be xAI, which effectively will say, ‘We’ll do whatever you want.'” The comment frames safety commitments as market positioning rather than absolute requirements—a tolerance cost OpenAI is willing to bear only if competitors face the same constraints.
The Congressional Battlefield
Altman faces a divided legislature. Senate Commerce Committee Ranking Member Maria Cantwell continues to advocate for “a more structured approach grounded in standards, testing and public infrastructure investment,” according to Holland & Knight analysis of White House AI policy. Democratic lawmakers including Senators Ed Markey, Brian Schatz, and Representatives Yvette Clarke and Don Beyer remain skeptical of federal preemption that would override state-level safety mandates.
Republican leadership, by contrast, has largely aligned with the White House’s deregulatory framework. House Speaker Johnson’s willingness to meet with Altman signals receptiveness to arguments that mandatory approvals would slow innovation and cede ground to China.
“We have to be very careful not to have these pre-approval requirements, including at state levels because that would really slow innovation in the country.”
— Sam Altman, OpenAI CEO
The Commerce Department’s Center for AI Standards and Innovation announced new pre-deployment testing agreements with Google DeepMind, Microsoft, and xAI on 5 May 2026, joining existing partnerships with OpenAI and Anthropic, HPCwire reported. CAISI Director Chris Fall described the work as “essential to understanding frontier AI and its national security implications.” But these agreements remain voluntary—the precise structure Altman is defending.
The China Paradox
Altman’s core argument hinges on competitive necessity: mandatory US approvals would hand advantage to less-regulated rivals, particularly China. Yet the empirical record complicates this narrative.
China has maintained the world’s most extensive AI regulations for the past four years, rolling out over half a dozen binding national requirements covering content censorship, psychological harms, job displacement, and mandatory labeling of AI-generated content, per the Carnegie Endowment for International Peace. During that same period, Chinese AI capabilities largely caught up with US technology.
Microsoft President Brad Smith testified in May 2025 that “the number one factor that will define whether the United States or China wins this race is whose technology is most broadly adopted in the rest of the world.” This framing shifts the competition from regulatory speed to international trust—a metric where safety commitments could prove advantageous rather than constraining.
The Chinese regulatory experience suggests that extensive oversight did not prevent technological progress, undermining claims that US pre-deployment reviews would necessarily create decisive competitive disadvantage. The real risk may be regulatory arbitrage: if US approval processes are slow or unpredictable, companies could prioritise international deployments while American testing lags.
The 500-Day Pivot
Altman’s current stance represents a notable shift from his May 2023 Senate testimony, when he advocated that “AI companies, especially those working on the most powerful models, adhere to an appropriate set of safety requirements, including internal and external testing prior to release and publication of evaluation results,” according to the Center for AI Policy. The language then suggested openness to mandatory frameworks.
By May 2025, that position had hardened into opposition to pre-approval requirements. The evolution tracks OpenAI’s commercial trajectory—early-stage caution when capabilities were uncertain has given way to deployment urgency as GPT-4 and GPT-5 generations prove commercially viable.
What to Watch
Congressional response in the next 30 days will determine whether Altman’s lobbying gains traction or faces organised Democratic resistance. Watch for:
- Bill language: Whether any Congressional AI legislation introduced in June includes mandatory pre-deployment review provisions or adopts the voluntary Commerce Department framework.
- Cantwell’s counter-proposal: Senate Commerce Democrats may draft competing legislation requiring CAISI testing before deployment, setting up a direct confrontation with OpenAI’s position.
- State-level moves: California, New York, and Massachusetts lawmakers continue developing their own AI safety requirements—Altman’s warning about “patchwork” regulation suggests OpenAI will seek federal preemption.
- International regulatory alignment: EU AI Act implementation proceeds with mandatory conformity assessments for high-risk systems. If European approvals become de facto global standard, US voluntary framework becomes irrelevant.
- Chinese capability benchmarks: Any evidence that Chinese frontier models match or exceed US performance would strengthen pro-regulation arguments that oversight didn’t prevent Beijing’s catch-up.
The real test comes in November’s midterm elections, where Leading the Future’s $125 million will target Congressional races. If the PAC successfully shifts chamber composition toward deregulation advocates, Altman’s June lobbying becomes a down payment on longer-term legislative control. If Beijing’s models match or exceed US capabilities during this window, it would validate the Carnegie Endowment finding that extensive regulation and technological leadership can coexist—directly undermining the innovation-versus-oversight framing that anchors OpenAI’s lobbying strategy.
The industry’s fracture is now complete. Anthropic has committed $50 million to tougher regulation and walked away from Pentagon contracts to preserve safety commitments. OpenAI has marshalled $125 million to elect lawmakers who will reject mandatory reviews. The battle is no longer about whether AI needs governance—both sides accept that reality. The fight is now over who controls the rules.