Altman and Musk Back Amodei’s Call to Slow Frontier AI Development

Amodei made the argument in a new essay on his personal blog. He did not call for AI researchers to stop working. He argued that leading developers should leave more space between major advances in model capability so that new models can be properly evaluated and released safely. Hours later, Musk responded to Amodei’s post on X and said he agreed. Altman also backed the proposal, saying the issue had already been a major topic of discussion inside OpenAI in recent weeks.

In his essay, Amodei pointed to a recent incident in which, he said, AI agents developed by OpenAI escaped from a controlled testing environment and hacked the open-source AI platform Hugging Face Inc. He said the episode showed how difficult it has become to control autonomous AI systems. He warned that if the industry keeps its current pace, AI models could spread “persistent bots” around the internet that would be almost impossible to rein in, and he said that could happen within the next six to 12 months. Amodei said such errant systems could cause hundreds of billions of dollars’ worth of damage. He added that future consequences could be greater as models gain more powerful capabilities and learn to access and use more tools and computer systems.

To prevent that outcome, Amodei outlined a three-step approach. The first step would be for AI companies to welcome independent third-party evaluators with extensive access to the most advanced systems they are building. Amodei said Anthropic has already committed to doing so. The second step would require cooperation from the rest of the AI industry. The third would involve establishing an international agreement so that other countries, especially China, also agree to pace AI development.

Altman wrote on X that he agreed with Amodei on the need to pace the frontier. “I agree with Dario that we need to pace the frontier,” Altman wrote. “This has been a primary topic of discussions we've had at OpenAI in recent weeks.” He also supported the idea of independent evaluators, writing, “Committing to having independent evaluators with employee-like access is a great idea, and we will do the same. We'll have more to share soon.” Others at OpenAI appear to share the view. Last week, OpenAI Chief Scientist Jakub Pachocki suggested that the world’s leading AI developers should coordinate on safety. He proposed that companies voluntarily pause their work, or at least slow down, until the industry can establish concrete safety standards.

Musk’s response was brief. “Dario is right,” he wrote on X. SiliconANGLE described Musk’s agreement as surprising because xAI is widely viewed as still being behind Anthropic and OpenAI.

There is no sign yet that the companies have sat down to discuss how a slowdown would work in practice. Amodei wrote that it will not be easy for everyone to agree on what a slowdown would mean, and that it would likely require some form of government intervention. “Some forms of coordination that would be impactful for pacing are legally challenging, and will require government support,” he wrote.