The UK’s AI Security Institute says Anthropic’s latest model has become the first AI system to complete a full end-to-end corporate network attack simulation, raising fresh concerns about the offensive capabilities of frontier AI models.
In a new thread on X, the AISI says Anthropic’s Claude Mythos Preview has completed a multi-stage network attack from start to finish.
“We conducted cyber evaluations of Claude Mythos Preview and found that it is the first model to complete an AISI cyber range end-to-end. The range simulates a 32-step corporate network attack, from initial reconnaissance to full network takeover. We estimate it would take a human expert 20 hours to complete.”
The institute says the result is a significant marker of how far AI capabilities have advanced in just a few years.
“In 2023, the best models could barely complete beginner-level cyber tasks. Today, our evaluation of Mythos Preview shows that it — and potentially future models — could be directed to autonomously compromise small, weakly defended, and vulnerable systems if given network access.”
AISI says the findings are an urgent call for stronger cybersecurity measures across organizations.
“These results underscore the importance of cybersecurity fundamentals like regular security updates, access controls, security configuration, and logging. Future models will be more capable still, but AI can also deliver advanced capabilities for defense.”
Last week, IBM warned that Mythos is capable of identifying multiple minor flaws and linking them together into a single, effective exploit, or what is known as “vulnerability chaining.” And last week, the Fed and the U.S. Treasury Department abruptly summoned the CEOs of the country’s largest banks to talk about the potential threats emerging from Anthropic’s most powerful AI model.
Disclaimer: Opinions expressed at CapitalAI Daily are not investment advice. Investors should do their own due diligence before making any decisions involving securities, cryptocurrencies, or digital assets. Your transfers and trades are at your own risk, and any losses you may incur are your responsibility. CapitalAI Daily does not recommend the buying or selling of any assets, nor is CapitalAI Daily an investment advisor. See our Editorial Standards and Terms of Use.