Mythos AI chains cyberattacks in UK security tests
TL;DR
UK's AI Security Institute finds Anthropic's Mythos Preview can autonomously chain a 32-step data-exfiltration attack in a simulated corporate network, outperforming prior models in the Last Ones test, while still failing tougher seven-step scenarios and real-world defense conditions—highlighting AI's potential to both threaten and strengthen cybersecurity.
Topics:businesstechnology#ai-security-institute#capture-the-flag#cyberattack#mythos#technology#the-last-ones
- UK gov’s Mythos AI tests help separate cybersecurity threat from hype Ars Technica
- OpenAI expands access to cyber AI as hacking risks grow Axios
- In the Wake of Anthropic's Mythos, OpenAI Has a New Cybersecurity Model—and Strategy WIRED
- Project Glasswing: Securing critical software for the AI era Anthropic
- Jamie Dimon says Anthropic's Mythos reveals 'a lot more vulnerabilities' for cyberattacks CNBC
Want the full story? Read the original reporting
Read on Ars Technica