Anthropic introduces powerful new AI model Mythos in the preview version of the brand new cybersecurity project.

Anthropic released a preview version of its new cutting-edge model, Mythos, on Tuesday, stating that this model will be available to a small number of partner organizations for cybersecurity-related work. In a leaked memo, the artificial intelligence startup referred to the model as one of their "most powerful" models to date. Anthropic said that the limited release of this model is part of a new security project called the "Glasswing Project," with over 40 partner organizations deploying the model for "defensive security work" and ensuring the security of critical software. The company mentioned that while Mythos is not specifically trained for cybersecurity tasks, this preview version will be used to scan code vulnerabilities in proprietary and open-source software systems. Anthropic claimed that over the past few weeks, Mythos has identified "thousands of zero-day vulnerabilities, many of which are high-risk." The company added that many of these vulnerabilities have existed for one to two decades. Mythos is a general model of Anthropic's Claude artificial intelligence system, which the company says has exceptional intelligence for programming and reasoning. Anthropic's cutting-edge models are considered its most sophisticated and high-performance model series designed for more complex tasks such as building and programming intelligent agents. Partner organizations participating in the Mythos preview test include Amazon, Apple, Broadcom, Cisco, Claudestra, Linux Foundation, Microsoft, and Palo Alto Networks. As part of the project, these partners will ultimately share their experiences using the model to benefit other entities in the tech industry. Anthropic stated that this preview version will not be fully open to the public. Anthropic also mentioned ongoing discussions with U.S. federal officials regarding the use of Mythos. However, it can be imagined that the related negotiation process is quite complex - previously, the U.S. Department of Defense classified the Anthropic AI lab as a supply chain risk due to Anthropic's refusal to authorize autonomous targeting or monitoring of U.S. citizens, leading to a legal dispute between Anthropic and the Trump administration.