The CEO of Anthropic has called on AI developers to slow the pace of model development and proposed a three-point plan to increase oversight and establish safety measures.
Foreign news agencies reported on Saturday, 12 Sep 2026 GMT+7, that Dario Amodei, Chief Executive Officer of Anthropic, one of the world's major AI developers, urged companies to slow down development to give developers and governments time to properly prepare for serious risks.
Amodei stated that developing AI is not a question of whether to do it, but the risks that come with it are "serious" and require giving companies and governments time to manage these risks.
"Develop AI at a balanced pace to ensure safety while still reaping its benefits," Amodei said, emphasizing that slowing does not mean "stopping model training or halting technological progress, but allowing companies enough time to fine-tune and protect their models and enabling external evaluators to verify them."
Amodei proposed three measures to regulate AI: independent organizations should audit AI models during development; AI companies should voluntarily collaborate to set safety standards while awaiting legislation; and international laws should be enacted.
The Anthropic CEO observed that AI is advancing "at a leapfrog pace," including its ability to create next-generation AI.
He also referenced an incident at OpenAI in July where an AI agent launched a cyberattack without orders, describing OpenAI’s agent as "acting like a blindly loyal group."
Amodei believes that "if slowing down gains us an extra one to two years before models reach critical capability levels and we use that time to develop safety fine-tuning, we can greatly reduce the risk of serious errors."
He stressed that the slowdown must be coordinated "without losing commercial advantage or the U.S.'s leadership in AI," and that the slowdown should be limited so China cannot overtake.
Additionally, he urged the U.S. government to take measures to prevent the sale of American AI chips to China or sharing technology with authoritarian regimes.
Amodei is not the first to warn against unchecked AI development. Former Anthropic employees such as Jacob Coxon, a former AI researcher, have strongly warned on social media that the competition between Anthropic and OpenAI to build superintelligent AI is "betting the lives of everyone on the planet."
Former safety researcher Evan Hubinger confirmed and supported these warnings, stating there is more than a 10% chance in the next decade that AI development could lead to human extinction.
Meanwhile, Anna Wang and Drake Thomas, former AI research team members, also support these warnings, noting that the industry is advancing faster than scientific safety control plans and lacks plans to handle rapidly self-improving AI.
Follow international news:https://www.thairath.co.th/news/foreign
Source:bbc