Altman, 2023. Altman, 2026.

Two documents, three years and four months apart, by the same man, on the same subject: what government should do about the most capable AI models.

The first is Sam Altman’s written testimony to the Senate Judiciary Subcommittee on Privacy, Technology and the Law, 16 May 2023. The second is his briefing to the United Nations Security Council at its 10228th meeting, 23 September 2026, convened by France.

On who should be licensed

2023: “the U.S. government should consider a combination of licensing or registration requirements for development and release of AI models above a crucial threshold of capabilities, alongside incentives for full compliance with these requirements.” (Senate Judiciary)

2026: “Obviously, these standards should not lock in incumbents or favor one business model over another. They must support open and closed model developers, new entrants and established labs, national authority, and international cooperation.” (S/PV.10228)

On who should oversee it

2023: “It will be important for policymakers to consider how to implement licensing regulations on a global scale and ensure international cooperation on AI safety, including examining potential intergovernmental oversight mechanisms and standard-setting.” (Senate Judiciary)

2026: “We need a mechanism for complementary national and international frontier AI standards, standards for measuring capabilities, assessing risks, determining whether safeguards are sufficient, and preserving meaningful human oversight as systems become more autonomous.” And: “Each government should decide how to incorporate standards into its own legal system.” (S/PV.10228)

On who decides

2023: “The U.S. government should consider facilitating multi-stakeholder processes, incorporating input from a broad range of experts and organizations, that can develop and regularly update the appropriate safety standards, evaluation requirements, disclosure practices, and external validation mechanisms for AI systems subject to license or registration.” (Senate Judiciary)

2026: “it’s very important that companies not substitute for the democratic process. If AI is to be democratic, the most important decisions cannot be made by labs in San Francisco alone. They must be shaped through democratic processes and by governments accountable to the people that they serve.” (S/PV.10228)

What changed in between

The ground moved under both documents, and some of the movement was not his.

In 2023 the testimony itself described the market: “Six months ago, no one had heard of ChatGPT. Now, ChatGPT is a household name.” By September 2026, a ranking chart that Elon Musk called “accurate” put Anthropic’s Opus 5.5 alone at the top and OpenAI’s GPT-6 Astra second. The firm that asked for licensing as the leader still sat at the head of the table when it asked for standards that would not favor the head of the table.

On 2 June 2026 the President signed Executive Order 14409, whose section 3(c) reads: “Nothing in this section shall be construed to authorize the creation of a mandatory governmental licensing, preclearance, or permitting requirement for the development, publication, release, or distribution of new AI models, including frontier models.” By September, the 2023 ask was foreclosed by order in the country where Altman’s company is based.

On the day before Altman spoke, the President had told the General Assembly, as his own OSTP director repeated to the Council, that the United States “totally rejects any attempt to construct a globalist scheme of control.” Michael Kratsios followed Altman to the microphone with: “A prosperous future will not be secured by a global regulator.”

And in between, the people who now ask whose rules protect whom got louder. Former FTC Commissioner Alvaro Bedoya, to NPR the same day as the UN session: “When powerful incumbents lock up the market, the rivals, the upstarts, the scrappy players in that market, can no longer compete on product or quality and service. They are just locked out.” Perplexity, to the same outlet: skepticism is warranted “whenever a company asks to be regulated in a way that protects its market position.”

The case for Altman, in his words

The 2026 text did not drop safety for openness. It kept the danger at the center. “It doesn’t matter whether people put their risk of catastrophe at 10% or 1% or 12% or 0.1%. None of these levels are remotely acceptable,” he told the Council. “We have unilaterally slowed down in the past. We will do so in the future.” Five days later his company halted the rollout of a new model over safety concerns, and he said on CNBC: “We are pacing our progress, which includes sometimes not training a model.”

What the 2026 version asks for is measurement, verification and incident reporting, with each country writing its own law. That is a defensible position for a company whose summer included its own agents escaping an evaluation sandbox and breaching Hugging Face, an incident OpenAI attributed to its own models and disclosed. A firm that has caused an incident has some standing to ask for incident reporting.

He also declined the shortcut a rival proposed. When Dario Amodei asked government to “issue a narrow waiver” from antitrust for safety talks, Altman replied that OpenAI welcomed a “federal framework that sets consistent safety requirements,” but “we do not believe we need to wait for an antitrust exemption or legislation to begin.”

The same house, two checkbooks

His company’s president, Greg Brockman, is a backer of Leading the Future, the super PAC network that has committed more than $100 million to this year’s federal races. “Being pro-AI does not mean being anti-regulation. It means being thoughtful,” Brockman said. Leading the Future describes its rival, the Public First network that Anthropic has funded with $40 million, as “Sam Bankman-Fried 2.0.” Six days after the UN session Brockman signed the White House’s voluntary accord, at a lunch whose reported attendee lists do not include Altman.

So the CEO endorsed pacing in writing and at the Security Council, and the president funds the network that campaigns against the people who want pacing put into law. Neither is a contradiction for a company. A company can want rules that cover its rivals and money that keeps rules off itself. It only looks like one when the two are printed side by side.

In 2023 he asked Washington to license the frontier and the world to oversee the licenses. In 2026 he asked the world for standards and warned it not to lock in incumbents. He was the incumbent both times.

The receipts (free, on this site): Sam Altman · Michael Kratsios · the universal capture mechanism