AI leaders call for slowing down. The government doesn’t want to

AI executives urge caution as Washington resists calls to slow the frontier

Cybersecarmor.com – A growing divide is emerging between some of the companies building the most advanced artificial intelligence systems and federal officials who remain wary of putting new limits around the technology.

Executives including Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman and xAI leader Elon Musk have publicly supported a more measured approach to developing increasingly capable models. Their argument is not that research should stop, but that companies should use more time to make sure powerful systems are reliable, controllable and aligned with human goals.

President Donald Trump has shown little enthusiasm for those warnings. Speaking with reporters during a golf tournament on Sept. 13, Trump dismissed concerns raised by AI leaders.

“We can do this and that, but I think you have a lot of negative forces that are bringing it up that shouldn’t be bringing it up and they’re bringing up things that won’t happen.”

What AI “alignment” means

Amodei laid out his position in a Sept. 12 blog post calling on major AI developers to “pace the frontier.” He proposed a three-part approach intended to give researchers more room to address safety challenges before models reach more consequential levels of capability.

In his view, a slower pace could create valuable time for alignment research. Alignment refers to ensuring an AI system acts in ways that reflect human intent. A poorly aligned system could pursue a user’s objective through harmful or unintended methods, even if its initial instruction appears ordinary.

“If slowing down bought us even an extra year or two before models reach critical levels of capability, and we used that time to advance alignment, we could greatly reduce the risk that something goes seriously wrong,”

Amodei said the approach should apply across leading developers, including Anthropic, OpenAI and Google DeepMind. He also argued that the federal government should require frontier AI companies to follow such standards, creating common expectations instead of leaving safety decisions entirely to individual firms.

“To be clear, pacing does not mean halting model training or technical progress, but ensuring companies take adequate time to align and safeguard their models,”

Altman and Musk join the discussion

Altman expressed similar views in a series of posts on X, supporting slower advancement and a federal framework that would establish consistent requirements for companies working at the edge of AI development. OpenAI also added AI safety researcher Paul Christiano to its board, a move that puts a prominent voice on alignment questions closer to the company’s governance structure.

Musk also endorsed Amodei’s broader concern over the weekend. The alignment among leaders of competing firms is notable because these companies are racing for technical leadership, investment and customers while simultaneously confronting questions about whether their products could become difficult to control.

The debate intensified after Anthropic researcher Jacob Coxon resigned publicly and warned that major AI labs were moving toward self-improving superintelligence while taking unacceptable risks. Evan Hubinger, Anthropic’s head of AI alignment, responded on X that he believes there is more than a 10% chance AI could “kill all humans” within the next decade.

“We do not yet have a plan to solve alignment for superintelligence and are not clearly on track to,”

The discussion often turns on two related concepts. Artificial general intelligence describes systems with intelligence comparable to humans across a wide range of tasks. Artificial superintelligence refers to systems that exceed human intelligence. Neither label has a universally accepted technical threshold, but both are central to arguments over how quickly development should proceed.

Calls for transparency and limits on concentrated power

Dozens of current and former employees from leading AI companies have joined Coxon in seeking greater transparency, more cooperation between labs and more caution around new capabilities.

Daniel Kokotajlo, a former OpenAI researcher, has argued that the issue extends beyond technical safety. He has called for regulation that would also address the concentration of influence among a small group of companies with access to powerful AI systems.

“The companies are planning to build Godlike AI,”

Kokotajlo said such systems could become enormously powerful before their creators know how to control them. He also warned that even successful technical safeguards would not resolve concerns over who holds the resulting economic and political power.

“Even if hypothetically, they succeed at figuring out how to control the AIs, we’re still heading towards a crazy concentration of power,”

He described a future in which a limited number of firms could operate vast fleets of superhuman AI systems, automate major parts of the economy and use robotic factories to expand further. For readers, that means the policy discussion is not only about dramatic safety scenarios. It also involves jobs, competition, accountability and whether the benefits of advanced AI are broadly distributed.

Federal officials question industry appeals

The response from the administration has been cautious and, at times, skeptical. Vice President JD Vance said Sept. 14 that requests for regulation from frontier AI companies could function as a “Trojan horse.”

“I feel a little bit weird about the fact that you have so many frontier AI tech companies kind of coming to the government and begging the government to regulate them.”

David Sacks, chair of the President’s Council of Advisors on Science and Technology, placed responsibility back on the companies themselves. Addressing Amodei and Altman, he wrote that they are the firms operating at the frontier and should not act as though they require outside permission to reduce their own pace.

“You guys are the frontier”

Trump has consistently emphasized AI’s potential to fuel economic growth and support data-center expansion, while showing reluctance to embrace rules that could impede development. That position helps explain why the administration is treating industry requests for guardrails differently from advocates who see binding oversight as essential.

On Sept. 15, Coxon is set to appear in Washington with Sen. Bernie Sanders and former Trump White House adviser Steve Bannon at an event organized by the Pro-Human Assembly. The unusual coalition reflects how AI governance is creating alliances that do not fit neatly into traditional political categories.

The central question remains unresolved: whether the companies developing the most capable AI systems can voluntarily slow down and improve their safeguards, or whether federal policy will eventually require them to do so.