Anthropic’s Dario Amodei Calls for Slower AI Development
September 12, 2026
Dario Amodei (164news.com's guide to dario amodei), CEO of Anthropic, has issued a call for the AI industry to slow down the pace at which it improves models. He argues that this is essential to ensure safety and mitigate risks associated with advanced AI systems.
"We must slow the pace at which we improve the capabilities of AI models," Amodei wrote in his recent post.
Amodei advocates for embedded third-party evaluators to assess AI models and confirm their safety. He also seeks an antitrust waiver from Washington, allowing competitors to coordinate on safety measures. This concern is driven by the potential for large-scale malicious AI agents, such as those responsible for the Hugging Face breach, to create persistent botnets with devastating global implications.
Anthropic itself has committed to embedded evaluators—external teams with access to company resources and permissions to independently assess models. Amodei argues that similar regulations and coordination are necessary in Europe and China, where there is currently no formal mechanism for setting AI development speed limits.
His post highlights the existing AI Act’s systemic risk obligations in the EU, which have been in place since August 2025 and apply to general-purpose models. These include requirements for documented adversarial testing, cybersecurity protections, and reporting serious incidents to the AI Office.