Anthropic Unveils Mythos AI Model Preview in Cybersecurity Initiative with Industry Partners

Anthropic released a preview of its advanced AI model, Mythos, on Tuesday as part of Project Glasswing, an initiative aimed at using AI to prevent cyberattacks facilitated by AI tools. The model has identified thousands of bugs and vulnerabilities in major software, including operating systems and browsers.

Anthropic is partnering with more than 45 organizations, including rivals like Apple and Google, to explore and advance AI-driven cybersecurity measures.

The company stated it is holding back the full release of Mythos due to risks it poses to cybersecurity. Project Glasswing focuses on testing the model's ability to detect and mitigate threats from AI-powered hacking. Collaborators will use the preview to evaluate how AI can strengthen defenses against faster and more sophisticated attacks.

Model Capabilities and Applications Mythos has demonstrated effectiveness in finding vulnerabilities across popular software programs.

CoinDesk reported that the model identified thousands of such issues in every major operating system and browser. Anthropic claims this positions Mythos as a tool for proactive cybersecurity. The initiative addresses concerns that AI advancements enable hackers to conduct attacks with greater speed and scale.

Engadget noted the prevalence of discussions on AI's negative impacts in cybersecurity, with Project Glasswing positioned as a countermeasure. The New York Times described the development as a potential reckoning for the field, highlighting the dual-use nature of powerful AI systems.

Industry Collaboration and Restraints Partners in Project Glasswing include technology giants and cybersecurity firms, totaling over 40 entities.

Wired reported that the group will leverage the Claude Mythos Preview to test capabilities against evolving threats. This collaboration aims to build collective defenses without immediate public deployment of the model. Anthropic's decision to limit access reflects broader industry caution.

The New York Times reported that the company is working with partners to assess safe applications before wider release. No specific timeline for full deployment has been announced.

Broader Implications for AI and Security The project underscores the growing integration of AI in both offensive and defensive cybersecurity strategies.

Systems from companies like Anthropic and OpenAI are accelerating attack capabilities, necessitating AI-based countermeasures. Anthropic emphasized that Project Glasswing seeks to ensure AI benefits outweigh its risks in sensitive areas like digital security.