AI Model Mythos Raises Cybersecurity Concerns Among Global Regulators

Anthropic's new AI model, Mythos, is a potential cybersecurity risk to banks due to its advanced coding capabilities. European and U.S. regulators are gathering information and discussing safeguards with industry leaders, as the model could exploit technological vulnerabilities at unprecedented scale and speed.

The European Central Bank supervisors are scrutinizing Anthropic's AI model Mythos, concerned about its potential to enhance cyberattacks, a source informed Reuters on Wednesday. Cybersecurity experts predict significant challenges for banks due to this AI's ability to exploit vulnerabilities in legacy systems, prompting regulators in the UK and the US to take notice.

Supervisors from the ECB are collecting detailed insights about the model, preparing to question banks under their supervision about readiness for this emerging risk source, according to an anonymous insider who wasn't authorized to speak publicly. Unlike the US, which has opted for direct meetings, the ECB will engage in its ongoing dialogue with bank officials.

While the ECB spokesperson offered no comments, Mythos' ability to high-level code poses unprecedented cybersecurity threats. Anthropic is not releasing it widely but initiated Project Glasswing for select evaluations. U.S. and UK authorities are warning financial sectors, with figures like Trump, Secretary Bessent, and Fed Chair Powell supporting AI safeguards to mitigate risks.

(With inputs from agencies.)