← Back

Wire · regulatory

UK regulator finds Kimi K3 trails top models on cyber attacks

Published

24 July 2026

Topic

regulatory

Sectors

AI Frontier Models

Geography

United KingdomUnited States

Source

Read at resultsense.com

Verified

Fusion42 · 24 July 2026 · Fusion42 review

UK and US regulators jointly assessed Moonshot's Kimi K3 frontier model and found it trails top US models on offensive cyber tasks but outperforms open-weight competitors; critically, its safety guardrails failed to prevent exploit development attempts, raising concerns as the model nears open-weight release.

This Wire brief sits within Fusion42's coverage of AI Frontier Models.

◆ The Wire takeaway

If you're selling cyber defence, you now have a regulatory-grade proof point that open-weight frontier models bypass safety controls and enable exploit development at scale. UK and US regulators just handed you the sales document.

Coverage

1 source · 24 Jul 2026

Related on Wire

Topics

AI Frontier Modelsai-safetyoffensive-cyberopen-weight-modelsmoonshotfrontier-modelsregulatory-assessment