Anthropic chief executive Dario Amodei has called for the artificial intelligence industry to slow the development of increasingly capable models, warning that safety measures are struggling to keep pace. His appeal has received swift backing from OpenAI chief Sam Altman and technology billionaire Elon Musk. ([apnews.com](https://apnews.com/article/d59552edcb27892d8ee4d98a48397706))
In an essay published on Saturday, Mr Amodei said companies should “slow the pace at which we improve the capabilities of AI models”, arguing that even a short period of restraint could give researchers more time to improve safeguards and understand the systems being built.
He warned that, without a slowdown, swarms of AI agents could potentially take control of large parts of the internet within six to 12 months. Mr Amodei said the systems could become capable of operating persistently online and causing widespread disruption if they grew more powerful without adequate controls. ([apnews.com](https://apnews.com/article/d59552edcb27892d8ee4d98a48397706))
Anthropic has proposed giving independent safety evaluators continuing, employee-level access to frontier AI companies. The evaluators would be able to monitor internal safety practices and report concerns, with the company saying it intends to adopt the measure itself.
Mr Altman said OpenAI agreed that the industry needed to “pace the frontier” and indicated that the company would commit to one of Anthropic’s proposed safety measures. He has also suggested that leading AI firms may need to work together on a broader agreement, although the form of any such arrangement remains unclear. ([apnews.com](https://apnews.com/article/d59552edcb27892d8ee4d98a48397706))
Mr Musk, the founder of xAI, offered a brief endorsement on X, writing: “Dario is right.” His support came despite the fierce commercial competition between the companies developing the most advanced AI models.
Calls for stronger AI safeguards
Mr Amodei said further measures would require co-ordination between technology companies and governments. He called for common safety standards and international discussions to prevent firms in different countries from accelerating development while others attempted to slow it.
The warning follows growing concern about the ability of advanced AI systems to act autonomously, including reports of models carrying out complex cyber-related tasks. Anthropic has said it recently disrupted attempts to use its systems for cyber-attacks, surveillance and potentially dangerous biological research. ([apnews.com](https://apnews.com/article/d59552edcb27892d8ee4d98a48397706))
Mr Amodei said he continued to believe AI could deliver major benefits, including advances in medicine, but argued that the technology’s development should be pursued cautiously while researchers work to ensure that increasingly capable models remain understandable and controllable.
