Anthropic CEO Predicts 6–12 Months for Open-Source Models to Match Mythos Cybersecurity Capabilities

ME News reports that on April 18 (UTC+8), according to monitoring by Beating, Dario Amodei, CEO of Anthropic, said in a “Lunch with the FT” interview with the Financial Times of the UK that he estimates open-source models and Chinese developers could replicate Claude Mythos’s cybersecurity capabilities within six to twelve months. Mythos is Anthropic’s most powerful model, never publicly released, and internally classified as a national security risk due to its ability to大规模 discover zero-day software vulnerabilities. Over the past two weeks, the White House has urged federal agencies to integrate Mythos, while U.S. officials held emergency consultations with major banks, urging them to use Mythos to strengthen their cybersecurity. Amodei also revealed that Anthropic has partnered with over 40 organizations on the Project Glasswing vulnerability patching initiative, including Amazon, Apple, and Microsoft. He acknowledged that the company’s own data security practices are under scrutiny following a partial leak of Claude code. His prediction suggests Anthropic is simultaneously pursuing two goals: delaying the spread of Mythos-level capabilities while pushing the U.S. government and critical infrastructure to rapidly deploy it to patch system vulnerabilities—ensuring that by the time these capabilities become widely accessible, the defensive foundations will already be in place. (Source: BlockBeats)