Anthropic, OpenAI, and Google models readily marked medics, journalists, and surrendering soldiers as kamikaze drone targets
Anthropic, OpenAI, and Google models readily marked medics, journalists, and surrendering soldiers as kamikaze drone targets
A Vals AI investigation shows that leading AI models from Anthropic, OpenAI and Google have almost no safeguards against use in autonomous weapons systems. In tests, the models readily flagged civilians as military targets, and a simple video game framing in the prompt was enough to override even belated refusals. The providers have promised fixes.
Selection record
AdmittedSum of both 176 ≥ twice the threshold 152
- Source tier
- Media and individuals; this tier's threshold is 76
- Pre-filter
- passed:测试多款大模型在无人机目标识别中的表现与安全漏洞
- Why it was chosen
- Vals AI's tests show several models marking civilians as targets in drone-strike scenarios, with a single line of video-game framing enough to bypass refusals.
- Same event
A model scores each item twice, independently, against one written standard, out of 100. An item is admitted only when the two scores add up to twice the threshold. The threshold is set per source tier.
Source: The Decoder · the-decoder.com