A month ago, OpenAI published “Industrial Policy for the Intelligence Age: Ideas to Keep People First”, its most detailed document to date on how superintelligence should be governed 1.
OpenAI acknowledges that developing superintelligence comes with severe risks: large-scale economic disruption, increased inequality, threats to social programs and democracy, mass unemployment, automated cyberattacks, criminal use of bioweapons, mass surveillance, etc. They also note that their models might escape their control. But they deem the disruption worth it, and the benefits to outweigh the risks.
The plan then asks governments to step in and impose “common-sense” regulation, public input on alignment, guardrails on government use of AI, and a tax system that shifts more weight onto capital to protect programs like Medicaid and Social Security from AI disruption.
It looks like OpenAI is taking risks seriously and asking to be regulated. But it has years of lobbying and political maneuvering pointing the opposite way. This post documents that pattern.
OpenAI Has Consistently Opposed Binding Regulation
OpenAI has claimed to be in favor of AI regulation since it was founded, while consistently working to prevent it. The clearest evidence is in how the company has responded each time a regulation has been on the table.
In 2022, the EU was finalizing the AI Act, which would put general-purpose AI systems under stricter requirements around transparency and oversight. OpenAI lobbied to weaken it. They argued that GPT-3, the state of the art at the time, was not a high-risk system and shouldn’t be treated as one. They also threatened to leave the European market if the Act wasn’t softened 2,3. This is from the company that was founded in 2015 on the premise that AI was so dangerous it shouldn’t be left to a profit-driven company. Altman’s pitch to investors and OpenAI’s early employees was, roughly, that only someone concerned enough about safety should build AGI, that this someone should be a non-profit, and that they would put safety first and support regulation 4. By 2022, “aggressively support” had become “lobby against.”
In 2024, California’s SB 1047 would have required safety testing for frontier models and given the state attorney general the power to sue companies for non-compliance when their systems posed a real threat. OpenAI opposed it, and lobbied Governor Newsom to veto it, which he did. Their stated reason was that AI should be regulated at the federal level and that SB 1047 would create an uncertain legal environment and endanger innovation. The bill called for third-party audits, incident reporting, safety protocols before deployment, whistleblower protections, and a public compute cluster for researchers. The new policy plan asks for “auditing regimes,” “incident reporting,” “mechanisms for public input,” and broader access to AI. The policy plan is now asking governments to do the things OpenAI killed when California tried to do them 5,6.
There’s a pattern across the tech industry that’s important to note here. Tech leaders donate to Trump, and shortly after, regulatory constraints on their companies ease. Nvidia got chip export controls relaxed. Apple got tariff exemptions. Amazon and Microsoft locked in multi-billion-dollar government services deals. Inaugural fund donations and super PAC contributions are common across the industry, and on their own none of them prove anything. But the correlation between donations and regulatory favors is consistent enough that individual donations can’t really be read in isolation anymore 7 1.
Later in 2024, Altman donated $1 million to Trump’s inaugural fund. On Trump’s first day back in office, he repealed Biden’s executive order that required AI companies to run safety tests and implement guardrails 8. Altman later publicly described Trump’s deregulatory approach as “a very refreshing change” 4. Greg Brockman, OpenAI’s co-founder and president, and his wife later donated $25 million to MAGA Inc., the pro-Trump super PAC 9. Trump then signed an executive order limiting states’ ability to legislate AI 10. Brockman praised the administration for its “growth-focused mindset” 11. OpenAI also obtained federal exemptions allowing it to build datacenters more quickly than standard environmental and permitting reviews would normally permit 12.
In 2025, the Brockmans helped launch and donated $50 million to Leading the Future, an organization that openly lobbies against AI regulation 13–15.
In 2026, OpenAI supported SB 3444, a state bill specifically designed to prevent AI labs from being held liable in case their systems cause critical harm 16,17.
OpenAI has a defence for each of these in isolation. The most charitable version is that the company prefers federal regulation to state regulation, and international coordination to either, on the grounds that fragmented rules are worse than coordinated ones. But OpenAI has also lobbied against the EU’s coordinated approach, and has not used its political access to push for the federal regulation it claims to want. In practice, the position “wait for better regulation” is unfalsifiable: every concrete proposal is the wrong one, and the right one is always somewhere else.
The pattern is that every time a binding regulation comes up, OpenAI opposes it, lobbies to weaken it, donates to politicians who block it, or supports alternatives that protect the company from accountability.
Beyond Regulation
The pattern isn’t limited to regulation. Almost every commitment in the policy plan has a matching behavior pointing the other way.
The plan warns about “power and wealth becoming more concentrated instead of more widely shared.” OpenAI was founded as a non-profit in 2015 because its founders thought a technology this powerful shouldn’t be owned by a profit-seeking company. Altman wrote at the time that the company would “obviously, comply with and aggressively support all regulations” 4. The non-profit structure was supposed to realize this vision. It was dismantled in 2024 under the exact kind of commercial pressure it was designed to resist 18. The policy plan now recommends that other AI companies adopt “Public Benefit Corporation” structures with “mission-aligned governance”, which is the same structure OpenAI moved to when it stopped being a non-profit.
The plan also warns that AI could “erode the tax base that funds core programs like Social Security, Medicaid, SNAP, and housing assistance,” and suggests rebalancing through “higher taxes on capital gains at the top.” Altman has financially backed and publicly praised an administration that is cutting Medicaid and doing the opposite of taxing capital 19.
The plan calls for guardrails on government use of AI and warns against “enabling pervasive surveillance.” Earlier this year, the Department of War designated Anthropic as a “supply chain risk to national security” because Anthropic refused to allow Claude to be used for fully autonomous weapons or mass domestic surveillance. The same day, the Pentagon announced a deal with OpenAI. OpenAI’s statement said that “two of our most important safety principles are prohibitions on domestic mass surveillance and human responsibility for the use of force, including for autonomous weapon systems” which are the exact points Anthropic had just been punished for holding 20,21.
The plan claims OpenAI has focused heavily on AI safety. The company’s Superalignment team was promised 20% of OpenAI’s compute. The New Yorker, reports that it received between 1 and 2 percent on aging hardware before being dissolved in 2024 4. The policy plan now asks the public sector to fund and coordinate the safety work that OpenAI said it would do internally 22.
There’s a more general point underneath all of this. The OpenAI leadership tells different audiences different things. To US intelligence officials, the message is that OpenAI needs billions in government support to stay ahead of China. To safety-focused audiences, the message is international coordination and careful development 4. To Trump the message is that deregulation is “a refreshing change”. To the readers of “Industrial Policy for the Intelligence Age,” the message is that the company wants to be regulated.
Sources
Footnotes
-
To be fair, this pattern isn’t really new or Trump-specific. Tech companies have made such donations to get their way around in previous administrations too. ↩