**UK gov’s Mythos AI tests help separate cybersecurity threat from hype**
The UK government’s AI Security Institute has published an initial evaluation of Anthropic’s Mythos Preview model, adding independent public verification to previous reports. While Mythos isn’t significantly different in terms of individual cybersecurity tasks compared to other models, its ability to effectively chain these into multistep attacks could set it apart. The AISI’s evaluations through Capture the Flag challenges show that Mythos can now complete over 85 percent of low-level tasks, marking a significant improvement from earlier models like GPT-3.5 Turbo.
**Takeaways:**
– **Initial Evaluation:** UK government’s AI institute publishes findings on Anthropic’s Mythos model.
– **Cyberchain Capability:** Highlights Mythos’ potential to perform multi-step attacks, distinguishing it from previous models.
– **Progression in Testing:** Shows steady improvement in AI security capabilities over time.
Originally published at arstechnica.com. Curated by AI Maestro.
Stay ahead of AI. Get the most important stories delivered to your inbox — no spam, no noise.