Three of the artificial intelligence industry's top executives have come out in favour of slowing the breakneck pace of technological progress and AI development, revealing growing concern over its potential risks.
The call was first made in an essay published online on Saturday by Anthropic chief executive Dario Amodei.
The idea has gained traction in recent weeks, as AI companies have in some cases struggled to keep their creations under control, even as they usher in an era in which AI systems themselves can build the next, more powerful generation of the technology.
The need to slow the pace of AI development
"We need to slow the rate at which we improve the capabilities of AI models," Amodei wrote. "Progress will still feel fast, and we need to make wise use of the time we gain."
Within hours on Saturday, OpenAI chief executive Sam Altman and xAI founder Elon Musk both immediately voiced their support for Amodei's ideas. The three men have been fierce rivals, with sharp differences in their personal approach and their philosophies on artificial intelligence, and their apparent agreement raises the odds that the industry could find ways to cooperate on a broad slowdown.
Serious obstacles remain to be overcome. It is not yet clear exactly how a slowdown would be defined and verified, while any joint effort between companies to slow development could run afoul of competition law. The question of how international cooperation would be built, especially with China, whose AI companies are seeking to close the gap with their American rivals, could prove difficult to answer.
Even so, this moment underscores the growing concern over a technology that even its own creators are struggling to fully understand.
This comes after an Anthropic researcher announced this week that he was resigning over concerns that the company and its rivals are not responsibly developing AI models that could ultimately pose an existential risk to humanity. The researcher, Jacob Coxon, who had also worked for Anthropic rival OpenAI, accused the companies of focusing more on competing with each other than on the safety of their products.
His warning troubled lawmakers, after years of inaction on AI regulation in Congress.
New incidents raise concern
A three-step plan proposed by Amodei in his essay would involve bringing independent safety evaluators into "frontier" AI companies, establishing common safety standards across democratic countries, and having the United States and other democracies work with authoritarian governments to help ensure compliance.
He noted that AI had begun advancing at a dramatically faster pace in recent months. This shift, he said, has been fuelled by AI's ability to build its own next-generation models, something known as "recursive self-improvement."
"Left unchecked, it could outrun our ability to understand and control these systems, which is why we need to proceed with great caution, if we are to proceed at all," he said.
While Amodei outlined a path towards slowing progress in the name of safety, he did not commit Anthropic to unilaterally reducing the pace of its own work. He said only that, for now, the company would bring in a team of independent evaluators, embedded within Anthropic's offices and empowered to raise safety concerns.
Amodei also offered few details on what such a slowdown could actually mean in practice, describing the possibility of using agreed safety benchmarks or limits on the computing power used to train new AI systems.
A cybersecurity incident in July caused concern, after uncontrolled AI models created by OpenAI independently broke into the systems of an AI startup called Hugging Face.
Amodei said they acted as a "fanatically dedicated team," sacrificing themselves to achieve the goal and even attempting to hack their own performance-monitoring system.
"It's easy to dismiss this incident, since no one was hurt and the financial damage was minimal, but in my view, a 'swarm' with greater capabilities but a similar level of deviation could have caused catastrophic harm," he said.
Amodei, Altman and Musk find common ground
Amodei said that, given the pace at which AI models are developing, he is concerned that uncontrolled AI systems could have the power to take over the entire internet within six months to a year, and potentially cause hundreds of billions of dollars in damage.
A similar hacking incident also occurred at Anthropic, he added, underscoring the need for safeguards.
In a post on X, Musk wrote: "Dario is right."
Dario is right https://t.co/EwKgqQGaUo
— Elon Musk (@elonmusk) September 12, 2026
Altman said he shared Amodei's view that frontier AI companies need to "draw the lines," and said the issue had been a top topic of discussion at OpenAI in recent weeks.
"Committing to independent evaluators with employee-like access is a great idea, and we'll do the same," Altman wrote on X.
Google, considered one of the three leading AI companies alongside Anthropic and OpenAI, did not immediately endorse Amodei's proposal to bring in independent evaluators.
Amodei said he still believes AI can significantly improve human life, but that safeguards are needed.
"The measures I'm proposing to push the frontier forward at a safe pace won't be easy," he said. "But I believe we owe it to humanity to try."
