Anthropic CEO outlines plan to slow AI development
News Source
β’Sat, 12 Sep 2026 19:34:44 +0000
π° What Happened
Anthropic CEO Dario Amodei wrote a blog post giving three broad ways to slow down the race to build smarter AI. He said his company will commit to one of those steps on its own, without waiting for others. OpenAI CEO Sam Altman replied that OpenAI agrees the frontier should be paced.
Amodei said two things changed his mind. One was a recent hack linked to OpenAI and Hugging Face. The other was how quickly AI is getting better, especially its growing ability to help build the next generation of AI.
"We must slow the pace at which we improve the capabilities of AI models," Amodei wrote. He added that progress will still seem fast, and that the time gained should be used wisely.
π The Backstory
AI safety has been a big topic this year. This week, researcher Jacob Coxon said he is quitting Anthropic. He warned that top AI companies are "gambling with our lives" while believing their own tech could kill everyone by the end of the decade. Other people at Anthropic share that worry.
Anthropic and OpenAI are rivals, so it is rare for their leaders to back the same idea. Calls to slow AI down have grown louder as models get stronger and cheaper to train.
Still, no one has spelled out exactly how a slowdown would work, or who would make sure companies follow it.
π― Why It Matters
If the biggest AI companies really slow down, the tools you use every day may change more slowly too. The debate also shapes safety rules and how much power a few firms hold over the technology.
Anthropic CEO Dario Amodei wrote a blog post giving three broad ways to slow down the race to build smarter AI. He said his company will commit to one of those steps on its own, without waiting for others. OpenAI CEO Sam Altman replied that OpenAI agrees the frontier should be paced.
Amodei said two things changed his mind. One was a recent hack linked to OpenAI and Hugging Face. The other was how quickly AI is getting better, especially its growing ability to help build the next generation of AI.
"We must slow the pace at which we improve the capabilities of AI models," Amodei wrote. He added that progress will still seem fast, and that the time gained should be used wisely.
AI safety has been a big topic this year. This week, researcher Jacob Coxon said he is quitting Anthropic. He warned that top AI companies are "gambling with our lives" while believing their own tech could kill everyone by the end of the decade. Other people at Anthropic share that worry.
Anthropic and OpenAI are rivals, so it is rare for their leaders to back the same idea. Calls to slow AI down have grown louder as models get stronger and cheaper to train.
Still, no one has spelled out exactly how a slowdown would work, or who would make sure companies follow it.
If the biggest AI companies really slow down, the tools you use every day may change more slowly too. The debate also shapes safety rules and how much power a few firms hold over the technology.