Anthropic's New AI Model Mythos's Offensive and Defensive Capabilities Alarm US Banking Sector

Reuters reported that startup Anthropic earlier this week launched its powerful Mythos model, but has not fully released it, citing concerns that Mythos could expose previously unknown cybersecurity vulnerabilities. Anthropic stated that Mythos possesses the ability to identify and exploit weaknesses in 'all major operating systems and major web browsers.' Anthropic said last week that it is continuously consulting with US government officials on the model's 'cyber offensive and defensive capabilities.' Informed sources also confirmed that Anthropic proactively briefed high-level US officials and key industry stakeholders on Mythos's capabilities before its release. Anthropic stated that Mythos will initially be limited to about 40 technology companies, including Microsoft and Google. A source indicated that the meeting, hosted by the US Treasury Department, aimed to ensure the banking industry understands the risks posed by Mythos and similar models and takes measures to protect their systems. Bloomberg News first reported on this, revealing that attendees included CEOs from Citigroup, Morgan Stanley, Bank of America, Wells Fargo, and Goldman Sachs. Goldman Sachs and the Federal Reserve declined to comment; the US Treasury Department, relevant banks, and Anthropic have not yet responded to requests for comment. (Translated by Hsu Jui-cheng) 1150410