Anthropic’s cutting-edge AI model, Claude Mythos Preview, has achieved a major breakthrough in cybersecurity by discovering thousands of vulnerabilities across major operating systems and web browsers. The company has opted to share the model with key organizations responsible for maintaining the internet’s infrastructure, including Amazon Web Services, Apple, and Google, as part of its Project Glasswing initiative.

The model’s capabilities have advanced significantly, enabling it to identify and exploit previously unknown vulnerabilities, including a 27-year-old bug in OpenBSD and a 17-year-old remote code execution vulnerability in FreeBSD. According to Nicholas Carlini from Anthropic’s research team, the model can create complex exploits by chaining together multiple vulnerabilities, making it a powerful tool for both cybersecurity and potential misuse.

Given its capabilities, Anthropic has chosen not to make Claude Mythos Preview publicly available, citing concerns over the potential consequences of its misuse. Newton Cheng, Frontier Red Team Cyber Lead at Anthropic, stated that the company is committed to deploying the model safely and responsibly, and is providing up to $100 million in usage credits and $4 million in direct donations to open-source security organizations to support this effort.

This decision underscores the importance of responsible AI development and deployment, as the potential fallout from the misuse of such powerful technology could be severe. As Anthropic’s experience demonstrates, the rapid progress of AI capabilities demands careful consideration and planning to ensure that these technologies are used for the greater good.

Photo by AS Photography on Pexels
Photos provided by Pexels