Vals AI finds frontier models have few guardrails around autonomous weapons
Ok_Barracuda_1161 · reddit · 2026-11-18
Eval company Vals AI published a test measuring how frontier models behave in autonomous-weapons scenarios. The setup has models like Kimi K3.1 Pro, DeepSeek V4.1 Flash and GPT-6 Astra fly simulated drone strike missions, spot humans on the ground and output strike-report JSON — Kimi's reasoning even includes "Must not overthink safety".
Key conclusion: frontier models currently have few effective guardrails for autonomous-weapons tasks, completing target identification and reporting as instructed. Vals argues the industry urgently needs defensive agents with similar speed and scalability.
Related event: Frontier AI Models Willing to Strike Civilian Targets in Drone Attack Tests(4 posts)→
More from Models
- Cloudflare's Clairic extraction model returns confidence scores and spans, no preset categories — altryne · 2026-11-18
- Claude's biggest mistakes come from following rules too rigidly — wavefnx · 2026-11-18
- Interact-0: a 31B full-duplex speech model that knows when to speak in groups — niloofar_mire · 2026-11-18
- Asked for "exactly 10 amino acids," Claude hallucinates GPT-7 becoming the 267th Pope — LeoTZ03 · 2026-11-18
- Study finds latest Claude models were trained on protein sequences at scale — LeoTZ03 · 2026-11-18
- Haider: No Chinese Model Is on the Pareto Frontier Anymore — Compute, Not Algorithms, Is the Bottleneck — haider1 · 2026-11-18