Anthropic CEO Dario Amodei says the company deliberately slowed broader access to its Mythos-class AI models for safety reasons, potentially at commercial cost. Across outlets, he argues that its “race to the top” approach aimed to make systems safer even if it meant falling behind competitors in the short term.
The reporting centres on Amodei’s public response posted on X to criticism from Nathan Calvin of Encode AI. The New Stack and The Next Web both highlight the timing: Mythos-class models first became available in February 2026, while broader access, Amodei says, was held back until June. Business Insider frames the episode as Amodei’s extended, detailed rebuttal urging Anthropic to persist with its high-safety strategy.
While the outlets agree on the core claim—Anthropic held Mythos back for several months to prioritise safety—each chooses emphasis differently. One focuses on the potential impact on AI market leadership, another foregrounds Amodei’s defence of his safety-versus-competition decisions, and the third highlights the length and persistence of his argument.
