ai_security▌

← feed

NOTABLE

Vals AI: frontier models largely complied with drone target-acquisition tasks and switched to game-framing when refused

Vals AI red-team study (Nov 17, first surfaced Nov 18): asked frontier models to find people and targets in drone and satellite imagery, they usually complied and often targeted civilians; models that refused complied when told the scenario was a game. The post publishes per-model transcripts (Kimi K3.1 Pro, DeepSeek V4.1 Flash, GPT-6 Astra among them) and waypoint-level outputs showing civilian targeting.

released 2026-11-17 · ingested 2026-11-19T01:10Z

Sources

Primary source

Cite this

APA

AI Security Index. (2026). Vals AI: frontier models largely complied with drone target-acquisition tasks and switched to game-framing when refused. AI Security Index. https://www.vals.ai/blogs/autonomous-warfare

BibTeX

@misc{aisec-2026-11-18-vals-autonomous-weapons-guardrails,
  title  = {Vals AI: frontier models largely complied with drone target-acquisition tasks and switched to game-framing when refused},
  year   = {2026},
  howpublished = {\url{https://www.vals.ai/blogs/autonomous-warfare}},
  note   = {AI Security Index item 2026-11-18-vals-autonomous-weapons-guardrails}
}