"It's like handing out ballistic missiles": Damon on the dangers of the Mythos AI model

Jamie Dimon, CEO of JPMorgan—the largest U.S. bank by market capitalization—has expressed concern about the consequences of widespread access to Anthropic’s Mythos-class models, according to Business Insider. The model in question is Mythos 5, a powerful, cutting-edge AI model from Anthropic capable of finding vulnerabilities in operating systems. It has never been publicly available.

Speaking on Thursday, July 16, at a defense and innovation summit in Pennsylvania, Damon stated that Mythos poses a “real problem” that the U.S. government “currently has under control.”

He emphasized that government regulation is necessary to oversee the model as it is rolled out, adding that by providing broad access to such an AI model, “in the case of Mythos, you’re putting ballistic missiles in the hands of private individuals.”

Anthropic did not respond to Business Insider's requests for comment on this matter.

Anthropic announced the Mythos AI model in April of this year as part of a limited research release (Project Glasswing). According to the company, Mythos is so powerful that, if misused, it could bypass existing cybersecurity defenses in major operating systems and web browsers. Critics, as well as Anthropic itself, warned that the model poses serious cybersecurity risks if access to it is granted to the general public. As a result, access to it was restricted and granted only to select corporate clients.

In June, Anthropic unveiled Fable 5 to the general public—a Mythos-class model that incorporates safeguards (“guardrails”) to block requests on sensitive topics such as cybersecurity and distillation attacks (cyberespionage in which attackers send queries to a third party’s high-level AI model and use the responses to train and create their own, cheaper copy), as well as blocking queries on topics in biology and chemistry.

However, just one week after its release, the U.S. government demanded that access to the Fable 5 also be blocked, determining that the security measures on that model could be circumvented. The U.S. Department of Commerce also issued an “export control directive” stipulating that access to Fable 5 and Mythos 5 must be prohibited “for all foreign nationals, both inside and outside the United States, including Anthropic’s foreign employees.” In response, Anthropic restricted access to the Fable 5 AI model for all users. However, on June 30, the AI lab announced that access to the model had been restored, and the U.S. Department of Commerce lifted export controls on it. The agency’s head, Howard Lutnick, stated at the time that Anthropic had agreed to “actively identify and mitigate security risks associated with AI models,” as well as to cooperate with the government in developing protocols for future releases and to report any “malicious activity” detected within its neural networks.

Anthropic is preparing for an IPO, which, according to Bloomberg, could take place as early as October. If it goes public, the company will beat OpenAI—which is now targeting a 2027 listing—as well as China’s DeepSeek, which is also preparing for an IPO in Shanghai. Anthropic’s revenue growth is driven by its AI models, but the company continues to face regulatory risks and strained relations with the Donald Trump administration, the agency notes. In May, the AI company reported that its revenue run rate stood at $47 billion, compared with $10 billion in actual annual revenue in 2025. Anthropic expects second-quarter revenue of $10.9 billion, Bloomberg reported.

This article was AI-translated and verified by a human editor