Altman Senate Testimony, May 16, 2023: AI Regulation

---

The Setting

The hearing was chaired by Senator Richard Blumenthal (Democrat, Connecticut), who chaired the Senate Judiciary Subcommittee on Privacy, Technology, and the Law. Blumenthal opened with a dramatic gesture: an AI-generated audio clip of his own voice, delivering an opening statement that he had not actually spoken. The clip was generated by a voice-cloning system, and it was designed to demonstrate the capabilities — and the risks — of the technology the subcommittee was examining. The demonstration was, by most accounts, effective — it gave the senators and the audience a concrete, visceral sense of what voice cloning could do, and it set the tone for a hearing that was more substantive and more engaged than most technology hearings in Congress.

The ranking member was Senator Josh Hawley (Republican, Missouri). Hawley framed the hearing in terms of national competition: he argued that AI was a battleground between the United States and China, and that the regulatory approach needed to account for the geopolitical stakes. The framing was, in some ways, a preview of the broader geopolitical conversation about AI that would intensify over the following years — a conversation that pitted those who saw AI primarily as a competitive race against those who saw it primarily as a governance challenge.

The hearing was held in the wake of several significant AI developments. ChatGPT had been released in November 2022, and it had become the fastest-growing consumer application in history. GPT-4 had been released in March 2023. The Future of Life Institute’s pause letter (see B41) had been published in March 2023, calling for a six-month moratorium on AI training. The CAIS Statement on AI Risk (see B38) would be published two weeks later, on May 30, 2023. The regulatory conversation was, by most accounts, the most active it had ever been, and the Altman hearing was its centrepiece.

---

The Witnesses

Altman was joined by two other witnesses. Christina Montgomery, the Chief Privacy and Trust Officer at IBM, represented the incumbent enterprise-tech perspective. Montgomery argued against overly burdensome licensing, warning that a heavy-handed regulatory approach could stifle innovation and disadvantage smaller companies. She advocated for a more targeted approach — focusing regulation on the highest-risk applications, rather than imposing a blanket licensing requirement on all AI systems. The IBM position was, in some ways, the traditional technology-industry position: regulate the risks, but do not regulate the technology itself.

Gary Marcus, the NYU professor and persistent LLM critic, argued for stricter regulation and independent auditing. Marcus, who had been a vocal critic of the AI industry’s hype and a sceptic of large language models’ capabilities, used his testimony to argue that the technology was not as mature as its proponents claimed, and that the risks — from bias to misinformation to economic disruption — were greater than the industry acknowledged. He called for independent auditing of AI systems, for mandatory transparency requirements, and for a regulatory framework that was more stringent than what Altman was proposing.

The contrast between the three witnesses was, in some ways, the point. Altman represented the frontier labs — the companies building the most capable AI systems. Montgomery represented the incumbent technology companies — the companies that had been using AI for years, in more limited forms. Marcus represented the academic and critical perspective — the view that the technology was oversold and the risks were undersized. The three perspectives, taken together, gave the senators a range of views to draw on, and the hearing was, by most accounts, more substantive than most Congressional technology hearings.

But the hearing was dominated by Altman. He was the reason the room was full, the reason the cameras were rolling, and the reason the hearing was being covered by every major news outlet. Altman, at thirty-eight, was the CEO of the company that had built ChatGPT, and he was, in May 2023, the most visible figure in the AI industry. His testimony was, by most accounts, the most consequential Congressional appearance by a technology CEO since Mark Zuckerberg’s testimony during the Cambridge Analytica scandal in 2018.

---

What Altman Said

Altman made several significant statements during the hearing:

On regulation. Altman called for the creation of a new federal agency to regulate AI — an agency that would license the most powerful AI systems and have the authority to revoke licenses if the systems were found to be unsafe. He compared the proposed agency to the International Atomic Energy Agency (IAEA) — the UN body that regulates nuclear technology — and he said that a similar approach was needed for AI. The comparison to nuclear regulation was, by most accounts, striking. Technology CEOs had, historically, opposed regulation — they had argued that regulation would stifle innovation, that the government did not understand the technology, and that self-regulation was sufficient. Altman’s willingness to embrace regulation, and to compare AI to nuclear technology, was a significant departure from the industry’s traditional posture.

On risks. Altman said he was “nervous” about the risks of AI. He acknowledged that the technology could be used to spread disinformation, to automate jobs, and to manipulate people. He said that his “worst fear” was that the AI industry would cause “significant harm to the world” before appropriate safety measures were in place. The admission was, by most accounts, notable. Technology CEOs did not, as a rule, publicly acknowledge the risks of their own technology, and they certainly did not describe their “worst fears” about it. Altman’s willingness to do so was, for the senators, a sign that he was taking the risks seriously — and it was, for the public, a signal that the risks were real.

On elections. Altman expressed particular concern about the use of AI in elections — the use of deepfakes, voice cloning, and AI-generated disinformation to influence voters. He said that this was the area where he was most worried about near-term harm. The concern was, in retrospect, prescient — the 2024 election cycle would see a range of AI-generated political content, from the New Hampshire Biden robocall to deepfakes in elections around the world (see B84).

On open source. Altman expressed caution about open-source AI, suggesting that the most powerful models should not be freely available because of the risk of misuse. This position was consistent with OpenAI’s closed-source approach, and it was contested by the open-source community, who argued that openness was essential for accountability and innovation. The debate about open-source AI would intensify over the following years, with the release of open-weight models like LLaMA, Mistral, and DeepSeek (see B12).

---

The Senators’ Response

The senators’ response to Altman was, by most accounts, unusually bipartisan and unusually engaged. In most technology hearings, the senators’ questions fell along partisan lines — Democrats focusing on consumer protection and bias, Republicans focusing on free speech and competition. The Altman hearing was different. The senators, regardless of party, seemed to share a genuine concern about the risks of AI, and they seemed genuinely interested in figuring out what to do about it.

Blumenthal, the chair, was broadly supportive of Altman’s regulatory proposals. He compared the moment to the early days of nuclear regulation, when the government had to build a regulatory framework for a technology that was too new and too complex for existing regulatory bodies to handle. The comparison was, in some ways, apt — AI, like nuclear technology, was a powerful, potentially dangerous technology that did not fit neatly into existing regulatory categories.

Hawley, the ranking member, was more sceptical. He pressed Altman on the competitive dynamics of the AI industry — on whether OpenAI’s support for regulation was, in part, a way to raise barriers to entry for competitors. The concern was, by most accounts, legitimate — companies that support regulation often do so because regulation makes it harder for new competitors to enter the market. Altman denied the charge, but the concern would persist, and it would become a central theme in the debate about AI regulation.

Other senators asked about specific risks — about bias, about job displacement, about copyright, about the impact on children. The questions were, by most accounts, more informed than the questions in most technology hearings, and they suggested that the senators had done their homework. The engagement was, for the AI industry, a sign that Congress was taking the technology seriously — and that regulation was, eventually, coming.

---

The Public Reaction

The public reaction to the hearing was, by most accounts, positive. Altman was widely praised for his willingness to engage with the regulatory conversation, and for his honesty about the risks. The coverage — in The New York Times, The Washington Post, The Wall Street Journal, and every major technology publication — emphasised Altman’s cooperative demeanour and his willingness to embrace regulation. The coverage also emphasised the contrast with previous technology-CEO hearings — particularly Zuckerberg’s — where the CEOs had been defensive, evasive, and reluctant to acknowledge problems.

The reaction was not, however, uniformly positive. Some critics argued that Altman’s embrace of regulation was a strategic move — a way to shape the regulatory framework in a way that favoured OpenAI, by raising barriers to entry for competitors. Others argued that the hearing was too focused on long-term risks (existential risk, AGI) and not focused enough on near-term harms (bias, discrimination, job displacement, copyright). The critics noted that the witnesses who would have been best positioned to speak about these near-term harms — the researchers, the affected communities, the advocacy groups — were not at the witness table.

---

What It Meant

The hearing was significant for several reasons. First, it was the first time a major AI CEO had publicly called for federal regulation of AI. Altman’s endorsement of a licensing agency was a notable departure from the technology industry’s traditional opposition to regulation, and it gave political cover to senators who wanted to act. The endorsement was, in some ways, the most consequential thing about the hearing — it shifted the Overton window of the regulatory conversation, making it acceptable for politicians to propose AI regulation without being accused of being anti-innovation.

Second, the hearing established Altman as the AI industry’s most visible public face. His calm, cooperative demeanour contrasted with the confrontational style of other technology CEOs who had testified before Congress, and it made Altman a sympathetic figure — the AI CEO who was willing to admit that his technology was dangerous and who was asking for help to manage the risks. The positioning was, by most accounts, deliberate — Altman and OpenAI had been cultivating this image for months, and the hearing was its most prominent public expression.

Third, the hearing led to a series of follow-up actions. Senator Chuck Schumer, the Senate Majority Leader, organised a series of “AI Insight Forums” — closed-door meetings between senators and AI industry leaders — beginning in September 2023. President Biden signed an executive order on AI on October 30, 2023 (Executive Order 14110, requiring companies to share safety test results). And the Bletchley AI Safety Summit was held in November 2023 (see B40), producing the Bletchley Declaration on AI safety.

---

The Aftermath

However, no comprehensive federal AI legislation has passed Congress as of 2025. The Schumer forums concluded without producing a bill, and the political dynamics of the US Congress have made it difficult to pass any significant technology regulation. The Biden executive order was rescinded by the Trump administration on January 20, 2025, and the new administration signalled a preference for less regulation, not more. The voluntary commitments (see B90) that followed the hearing were, as their name suggests, voluntary — and they proved to be as limited in practice as their critics had predicted.

The Altman hearing was, in retrospect, a high-water mark for the push toward federal AI regulation in the United States. It was the moment when the regulatory conversation was most open, most bipartisan, and most serious. Whether that conversation will be revived — and whether it will produce legislation — remains an open question. The lesson of the hearing, for the AI industry and for the government, is that the regulatory conversation is not a one-time event but an ongoing process — one that requires sustained engagement from both sides, and one that can be derailed by political change, by industry lobbying, and by the sheer difficulty of regulating a technology that is evolving faster than the legislative process can keep up with.

---

---

---

Was Altman’s Senate testimony inevitable — the product of forces too large to redirect — or was it a series of choices, each of which could have gone differently? The answer matters, because it determines whether the future is something that happens to us or something we make.