Anthropic recently unveiled Mythos, its most advanced AI model, designed for defensive cybersecurity tasks. The model has already uncovered "thousands" of major vulnerabilities across major operating systems and web browsers.
Launched via "Project Glasswing," Mythos preview access was granted to tech leaders like Amazon, Microsoft, Nvidia, and Apple, alongside over 40 other organizations maintaining critical software infrastructure.
Experts warn Mythos can identify and exploit unknown vulnerabilities faster than they can be repaired, potentially accelerating sophisticated cyberattacks, particularly in sectors like banking. Anthropic acknowledges the model's flaw-finding ability, if misused, poses risks to economies, public safety, and national security.
Following the April 7 launch, U.S. software stocks declined as AI's potential disruption to traditional firms resurfaced.
The White House has engaged with Anthropic CEO Dario Amodei regarding Mythos, discussing cybersecurity and balancing AI innovation with safety. Despite a Pentagon supply-chain risk designation on Anthropic, the U.S. government plans to make a version of Mythos available to federal agencies. U.S. Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell briefed major U.S. bank CEOs on potential risks. British authorities are also in talks with banks and cybersecurity officials. Banks are coordinating with European regulators on the potential impacts.