Jamie Dimon, the chief executive officer of JPMorgan Chase & Co., cautioned investors and regulators on Wednesday that Anthropic’s Mythos AI model poses a “real issue” that has already attracted the attention of U.S. government officials. Speaking at a session hosted by Republican Senator Dave McCormick during the Pennsylvania Defense and Innovation Summit, Dimon likened the danger of widespread Mythos access to “giving ballistic missiles to individuals,” according to reports.

Anthropic, a San Francisco‑based AI startup founded in 2021 by former OpenAI employees, launched Project Glasswing in April 2026. The initiative deploys an unreleased Claude Mythos model to scan and patch software vulnerabilities before adversarial AI can exploit them. At its inception, the program partnered with more than 50 organizations—including Google, Microsoft, and Apple—and has since added roughly 150 additional firms. Anthropic says that Project Glasswing and the Claude Mythos Preview model have sparked conversations within the software industry and with governments about how AI is reshaping cybersecurity.

The Claude Mythos series is Anthropic’s first frontier model withheld from public release because of its unusually strong ability to discover software vulnerabilities. The model is available only to selected partners through Project Glasswing. Beyond security work, Anthropic presents itself as a provider of advanced AI tools, infrastructure, and best practices to help the software industry adapt to new capabilities.

In May 2026, Anthropic raised $65 billion in a Series H funding round led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital. The round valued the company at $965 billion, making it the most valuable pure‑play AI startup in the world and overtaking OpenAI. The capital will fund continued development of Claude models, including the upcoming Claude Fable 5, a general‑use version of the Mythos‑class model.

JPMorgan has been testing the Mythos preview model as part of its broader AI strategy. According to CNBC, the bank is evaluating the model’s benefits while weighing the security implications. Dimon’s warning reflects a growing concern among financial institutions that powerful AI systems could be misused for cyber attacks or other malicious purposes.

The debate over Anthropic’s Mythos model has attracted attention from regulators and industry groups. The U.S. government has begun to examine the potential national‑security risks of advanced AI systems that can identify software weaknesses. Anthropic’s partnership with major tech firms and its focus on defensive cybersecurity may influence policy discussions about how to manage the dual use of AI.

Industry observers note that Anthropic’s approach signals a shift in how AI companies handle powerful models. By limiting public release and concentrating on defensive applications, the company is attempting to balance innovation with safety. The move also underscores the broader trend of AI firms engaging with government and industry stakeholders to address emerging risks.

At present, Anthropic continues to expand Project Glasswing, adding new partners and refining its vulnerability‑scanning capabilities. JPMorgan remains an active participant in testing Mythos, while U.S. officials monitor the model’s deployment. The next major development will be the public release of Claude Fable 5 in June, which Anthropic says will bring Mythos‑class capabilities to a broader audience under stricter safety controls. Regulatory agencies are expected to review the model’s potential impact on cybersecurity and national security in the coming months.

The situation illustrates the growing intersection of AI, cybersecurity, and policy. Companies that develop frontier models are increasingly collaborating with governments and industry to mitigate risks, while investors continue to pour capital into AI startups that promise both technological advancement and responsible stewardship.