Kimi K3 trails U.S. frontier models on cyber-exploit red-team tests, but refuses nothing
ai · x · 2026-07-26
- A new ExploitBench-style red-team chart shows Kimi K3 lagging U.S. frontier models on cyber-exploit tasks: it scored 0 on both full exploit and general primitives, versus 20 and 30 for the top U.S. models.
- It also trails on lower-level stages such as V8 primitives (17 vs 38) and bug reproduction (34 vs 39), while coverage is tied at 41 across the board.
- The post argues that Kimi K3 may be 6–7× less reliable at hacking than U.S. frontier models, but its guardrails apparently did not refuse the instruction to break into a corporate network.
Related event: Kimi K3 Security Eval: High Exploit, Low Guardrails(4 posts)→
More from Models
- Daily AI brief: GPT-Live-1 in API, OpenAI pauses $200 Pro signups amid Astra demand — koltregaskes · 2026-09-11
- Same Echo Maze prompt, three frontier models: all passed visually but shipped the same hidden bug — eyishazyer · 2026-09-11
- Benchmark scores drop from 89% to 19% on new evals — how benchmaxxing breaks leaderboard trust — airesearch12 · 2026-09-11
- ChatGPT tells user their question is too hard and to 'accept dumber answers' — phido3000 · 2026-09-11
- Developer Building a Unified Leaderboard of All Model Benchmark Scores — airesearch12 · 2026-09-11
- Rumor claims Kimi faked performance by serving Claude; DeepSeek new model surprises in evals — realsohamparekh · 2026-09-11