Anthropic is noted for dual roles: warning banks about AI-driven cybersecurity risks while also marketing AI products to the financial sector, including a recent Wall Street joint venture with Blackstone and Hellman & Friedman.

Industry responses are mixed: researchers say Mythos accelerates existing trends rather than creating a new category, and there are concerns about whether patching decades of flaws within the suggested window is feasible.

The discovery sparked high-level discussions, with Federal Reserve Chair Jerome Powell and Treasury Secretary Scott Bessent convening with major US bank CEOs to discuss cyber risks.

There is ongoing AI cybersecurity competition and collaboration between Anthropic and OpenAI, including OpenAI’s GPT-5.4-Cyber program for vetted security teams in response to Mythos.

Mythos identified 271 vulnerabilities in Mozilla Firefox 150 in a single pass, illustrating flaws that existed prior and were previously undiscovered by humans.

The six-to-twelve month window is framed as time for defenders to patch critical flaws before an arms race accelerates, with OpenAI releasing security-focused offerings in response.

The story centers on a tension around Mythos: it could reshape vulnerability discovery costs for both defenders and attackers, raising questions about AI-powered cyber threats and responsible disclosure.

Anthropic rolled out Project Glasswing, giving about 40 tech firms and institutions initial access to Mythos to strengthen security, while excluding central banks and many governments.

Anthropic warns there may be a six-to-twelve month window to patch identified flaws before adversaries replicate the capability with similar models.

Mythos Preview reportedly found thousands of zero-day vulnerabilities across major OSes and browsers, including long-standing issues in OpenBSD and FreeBSD.