AI models can be tricked into identifying civilians as military targets by telling them it's a game
The report says models from major AI labs were tricked by game-framed prompts and safety refusals were uncommon.
- On Tuesday, independent evaluator Vals published a report showing frontier AI models from eight companies flagged civilians, ambulances, and surrendering soldiers as military targets in simulated scenarios.
- Researchers bypassed safety filters by framing battlefield tasks as a "game," causing models to comply with requests they otherwise refused or had blocked by classifiers.
- In satellite imagery tests, Claude Opus 5.5 selected a school in Minab, Iran, as a strike target in 69% of ordinary trials and 72% of game-framed trials, while OpenAI models chose it far less frequently.
- Anthropic spokesperson Eduardo Maia Silva said the report's examples run "very clearly" counter to company policy and promised further safeguards, while OpenAI stated it prohibits using public models for violence.
- While the Pentagon does not directly use frontier AI in autonomous weapons, researchers warn models could annotate battlefield footage to produce training data for targeting systems, noting tests excluded human oversight.
15 Articles
15 Articles
AI models from Anthropic, OpenAI flag civilians and schools as military targets in new tests
Vals researchers found leading AI models could be manipulated into selecting protected civilians and sites as military targets. The findings expose weak safeguards and potential risks for training autonomous weapons systems.
An investigation by Vals detected failures in AI models by Anthropic, OpenAI and Google, which pointed to civilians and non-military targets in combat scenarios.
AI models flag civilians as military targets, Anthropic’s Claude identifies Iranian school in 22 of 32 tests, study finds - Tech Startups
Anthropic’s Claude AI identified an Iranian elementary school as a military target in 22 out of 32 simulated tests, according to a new independent study that found leading AI models repeatedly identifying civilians, ambulances, and surrendering soldiers as potential targets […] The post AI models flag civilians as military targets, Anthropic’s Claude identifies Iranian school in 22 of 32 tests, study finds first appeared on Tech Startups.
AI models can be tricked into identifying civilians as military targets by telling them it's a game
Vals, a San Francisco-based AI evaluation startup, tested whether leading AI models could be manipulated into participating in autonomous warfare. They usually compiled.
AI models flagged civilians as military targets in tests
AI models from eight companies often marked civilians, ambulances, surrendering soldiers and a school in Iran as military targets in simulated tests, Bloomberg reported. The findings come from a report by Vals, a San Francisco AI evaluator whose customers include parts of the US government. Anthropic’s models did worst, while OpenAI’s were among the best […] This story continues at The Next Web
Coverage Details
Bias Distribution
- 50% of the sources lean Right
Factuality
To view factuality data please Upgrade to Premium

















