Anthropic's Dario Amodei answers a safety critic point by point, conceding three mistakes as the company heads for IPO
What's going on? Anthropic CEO Dario Amodei has publicly answered one of his sharpest critics, point by point. Nathan Calvin, chief strategy officer at the AI policy nonprofit Encode, posted a nine-item list on X this week arguing that Anthropic's recent moves look most consistent with a company "primarily trying to win the race to super intelligence", with safety "fundamentally secondary". He called it "a deliberately one sided list" and asked: "When Anthropic strays and makes mistakes, why do the mistakes systematically tend to guard and favor Anthropic's interests?" Amodei replied on X in the early hours of 21 Oct with a roughly 2,000-word post: "I disagree with the thrust of your post."
What Amodei defended and what he conceded: Calvin's first four items (an in-house inference chip team, more than $500bn in compute deals, Claude leading a growing share of AI development work, and heavy internal use of compute for AI R&D) are "simply examples of what it means to compete at the frontier, and we stand by them," Amodei wrote. He called the rewrite of the Responsible Scaling Policy, which dropped unilateral pause commitments, "the most substantive item". He conceded the last three: a letter to Congress that was later walked back, the remark that "you can't do safety from second place", and the draft federal AI bill Anthropic has been pitching. He blamed fast-moving communication in a company of over 5,000 people, adding: "I am of course the CEO, and I take responsibility." Against Calvin's list he set nine decisions he says cost Anthropic money, including backing California's SB 53, strict bio safeguards that "biologists joke about", holding back Mythos-class models from February to June, chip export-control advocacy and the 5% public benefit fund.
Why it matters: Anthropic is heading for a public listing reportedly targeting a valuation of around $2tn, and its draft S-1 shows $16.3bn of first-half revenue against an $80.7bn loss. Its claim to be the "safety lab" is central to that equity story, so a frontier-lab CEO litigating his safety record in public with an NGO critic is unusual. The reaction was split. Vox's Sigal Samuel asked what the evidence is that the "race to the top" strategy works, saying that from the outside it looks like Anthropic has done more to drive a race to the bottom, while others pointed out that the draft bill does not give government the blocking power Amodei says he wants.
What's confirmed and what isn't: Both posts are public and quoted directly here. The cost and timing claims in Amodei's list (for example the February date for Mythos-class models) are his own account and have not been independently verified. The draft federal bill has circulated but not been published, and Amodei promised to look further into Calvin's point on Anthropic's frontier compliance framework and to "have more to say soon".
Sources: Business Insider / Techmeme (Amodei on X) / India Today / explainx.ai (both lists side by side)
