Anthropic Claude Mythos: New AI Model Sparks US Financial Alarm Over Security Risks
- Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell summoned Wall Street leaders to an urgent meeting on April 10, 2026, following concerns over the capabilities of...
- The meeting was triggered by the introduction of Claude Mythos Preview, an advanced AI model designed to identify security flaws and weaknesses within software.
- Anthropic announced the model on April 7, 2026, as part of a cybersecurity initiative known as Project Glasswing.
U.S. Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell summoned Wall Street leaders to an urgent meeting on April 10, 2026, following concerns over the capabilities of a new artificial intelligence model developed by Anthropic.
The meeting was triggered by the introduction of Claude Mythos Preview, an advanced AI model designed to identify security flaws and weaknesses within software. The potential for the model to be exploited by bad actors has prompted a high-level response from U.S. Financial regulators and the Treasury.
Project Glasswing and Limited Rollout
Anthropic announced the model on April 7, 2026, as part of a cybersecurity initiative known as Project Glasswing. Due to the risks associated with its advanced cyber capabilities, the company is limiting access to the technology rather than releasing it to the general public.
The model is being rolled out to a select group of approximately 40 companies intended for defensive security work. Initial launch partners include:
- Microsoft
- Amazon Web Services
- Apple
- Nvidia
- CrowdStrike
- Palo Alto Networks
Dianne Penn, Anthropic’s head of research product management, stated in an interview with CNBC that the company viewed this limited release as a first step to provide cyber defenders with a head start on an increasingly important topic.
Market Impact and Security Concerns
The announcement followed the discovery of model descriptions in a publicly accessible data cache by Fortune in late March 2026. This leak reported that the model possessed advanced cyber capabilities that posed a significant risk, which subsequently led to a decline in cybersecurity stocks.
Anthropic has characterized the model as a cybersecurity reckoning
, emphasizing its ability to pinpoint vulnerabilities that could be used for both defensive and offensive purposes.
The model, Claude Mythos Preview, excels at identifying weaknesses and security flaws within software and Anthropic is limiting access to try to prevent bad actors from exploiting that capability
Anthropic via CNBC
The ability of the AI to discover thousands of vulnerabilities has created a sense of urgency among U.S. Officials, leading to the emergency summons of bank CEOs to discuss the systemic risks the technology might pose to the financial sector.