UK gov’s Mythos AI tests help separate cybersecurity threat from hype
Summary
Ars Technica reports on the UK's AI Security Institute evaluating Anthropic's Mythos Preview. Mythos performs similarly to peers on individual cyber tasks but can chain steps to execute long infiltration sequences, achieving a high score on the 32-step The Last Ones test. The evaluation notes limitations and suggests using AI to strengthen defenses, while cautioning that well-defended systems may resist automated attacks.