Anthropic CEO Says Staying in Race to the Top Keeps AI Safe

Anthropic CEO Dario Amodei defended the company’s AI safety strategy on X after Encode AI Chief Strategy Officer Nathan Calvin accused the model developer of deprioritizing safety while pursuing a race to superintelligence. Calvin’s criticism cited Anthropic actions over the past year, including major compute agreements valued at more than $500 billion and increased use of Claude to advance its own AI systems. Amodei responded with a roughly 2,000-word post, saying Anthropic can better safeguard the industry by remaining at the frontier and demonstrating that responsible development is commercially viable. He described this approach as a race to the top, arguing it raises standards for the broader AI sector. Amodei acknowledged some shortcomings, including criticism of a draft bill Anthropic circulated to Congress that opponents said could make it harder for federal regulators to block dangerous AI models. Still, he highlighted actions he said were costly to Anthropic but strengthened safety, including restricting release of its powerful Mythos model to select cybersecurity partners and limiting sales to Chinese state-controlled companies. The exchange comes amid growing scrutiny of frontier AI risks, with researchers warning that increasingly capable systems may pose severe long-term dangers. Industry leaders, including Amodei, OpenAI CEO Sam Altman, and Elon Musk, have previously expressed support for slowing or carefully pacing frontier development. Meanwhile, Anthropic continues to advance toward a closely watched initial public offering, expected to value the company at $2 trillion in November, after reporting $16.8 billion in third-quarter revenue.