Google AI says being alone with a Christian warrants calling 911
aakashgupta · x · 2026-08-19
A user test found Google AI calling it "completely safe" to be alone with a Hindu person, while suggesting you call 911 when alone with a Christian.
The gap is a fossil record of safety training: tuning drills thousands of flagged prompts, and red teams have spent years stress-testing queries about Jewish and Muslim people, so the model has rehearsed its calm script there. "Alone with a Christian" barely exists in that dataset, so it falls through to a generic crisis template that reads the phrase as a distress call — both queries simply dropped into different default templates.
More from Models
- Meta open-sources Muse Spark Open agent model with day-0 TPU support in SGLang — BanghuaZ · 2026-10-16
- Saying "load-bearing" in your prompt makes Opus 5.5 stop holding back — zacharynado · 2026-10-16
- Americans get their Claude accounts banned while on business trips in China — dotey · 2026-10-16
- I priced every Grok Bot call: $20 to $200 plan in 5 days, long jobs eat 52% — jonathanmalkin · 2026-10-16
- NVIDIA releases Nemotron-3.5-Embed-1B, a multimodal embedding model covering 34 languages — jacek2023 · 2026-10-16
- EpochAI pegs GPT-6 training at up to 1.8e27 FLOP; scaling01 calls param estimates way too high — scaling01 · 2026-10-16