Anthropic is in discussions with the US government regarding its new Mythos AI model, which the company deems too powerful for public release due to unprecedented cybersecurity risks. The banking industry is also expressing significant concern.
Anthropic's co-founder stated on Monday at the Semafor World Economy event in Washington, "The government has to know about this stuff." He confirmed ongoing dialogue about Mythos and future models, emphasizing the company's commitment to national security despite past disputes over defense department usage of their AI.
Last week, Treasury Secretary Scott Bessent met with senior American bankers in Washington to discuss the Mythos model. This meeting encouraged banking executives to test the model for vulnerability detection. Anthropic plans a limited release to select tech and cybersecurity firms, including Amazon, Apple, and JP Morgan Chase. Goldman Sachs, Citigroup, Bank of America, and Morgan Stanley are reportedly testing the model as well.
The UK's AI Security Institute issued a warning that Mythos represents a significant advancement in cyber threat capabilities, and UK financial regulators are also discussing its potential risks.