SAN FRANCISCO, US- Anthropic Chief Executive Dario Amodei has called on artificial intelligence companies to slow the development of increasingly powerful models, warning that the technology could advance faster than humanity’s ability to control it.
Amodei said leading AI developers should deliberately moderate the rate at which they improve their models to create more time for governments, researchers and companies to address emerging safety risks.
“We must slow the pace at which we improve the capabilities of AI models,” Amodei wrote in an essay, adding that progress would remain rapid even under a more cautious approach.
The Anthropic chief proposed a three-step framework for managing the development of advanced AI, including stronger safety evaluations and cooperation among major laboratories.
Such coordination would help ensure that companies prioritising safety are not disadvantaged by competitors racing ahead.
His intervention comes amid growing concern that future AI systems could become capable of improving themselves without direct human involvement.
The process, known as full recursive self-improvement, could allow an AI system to develop increasingly capable versions of itself.
Researchers fear that humans could struggle to understand, supervise or stop such systems once the process begins.
Anthropic co-founder Jack Clark previously compared the challenge to building a powerful car with an effective accelerator but no reliable brake. He said the industry needed a global mechanism for slowing development if dangerous capabilities emerged.
According to CNN, Anthropic believes a global response would be necessary because a decision by responsible companies to pause could allow another company or country to move ahead.
Amodei has previously estimated there is a 10 to 25 per cent possibility of AI producing an outcome that goes catastrophically wrong for human civilisation.
He said the danger could arise from the behaviour of the models themselves, misuse by individuals or governments, conflict between countries, or society’s inability to manage the technology.
However, he has also maintained that AI could deliver major benefits if developed safely, particularly in medicine, science and economic productivity.
The warning comes as Anthropic, OpenAI, Google and other technology companies invest heavily in developing more capable AI systems. Competition between the United States and China has also increased pressure on companies to advance quickly.
Amodei’s proposal seeks to give developers more time to understand and control increasingly autonomous systems before they reach capabilities that may be difficult to reverse.
“Progress will still seem fast, and we must make wise use of the time we gain,” he said.
The call places Anthropic among the leading AI companies publicly acknowledging that the race to build more powerful systems could carry risks that the industry is not yet fully prepared to manage.
