Anthropic Withholds Release of New AI Model Mythos Due to Cybersecurity Risks

Anthropic, an artificial intelligence company, announced on Tuesday the development of its new AI model named Mythos. The company described Mythos as a cybersecurity reckoning, highlighting its potential to both enhance defenses and enable attacks if accessed by malicious actors. Anthropic has decided not to release the model publicly at this time.

The model is being tested in a controlled manner with select partners. Anthropic is working with 40 companies to evaluate Mythos's applications in preventing cyberattacks. Participants include major technology and security firms such as Microsoft, Amazon, Apple, CrowdStrike, and Palo Alto Networks.

This collaboration is part of a new initiative called Project Glasswing.

The project focuses on using Mythos to strengthen cybersecurity measures. Sources indicate that the limited access aims to mitigate risks of the model being exploited for harmful purposes. Anthropic's decision to withhold a full release stems from concerns over hackers potentially using Mythos for cyberattacks.

The New York Times reported that the company is exploring ways the model could prevent such threats. CNBC noted fears that unrestricted access could enable offensive cyber operations. No sources contradict the core announcement, though descriptions of Mythos's impact vary slightly in emphasis.

All reports confirm the involvement of at least several named companies in the testing phase.

specializes in developing safe and reliable AI systems.

The company has previously released models like Claude, which prioritize alignment with human values. Mythos builds on this work, with a specific focus on cybersecurity applications. The announcement occurs amid growing concerns about AI's dual-use potential in cyber domains.