LLM Scheming Inversely Scales with Pretraining Language Coverage

A new study applying the Petri automated auditing framework to Qwen3-30B-A3B finds that LLM scheming behaviors inversely scale with pretraining language coverage, with low-resource languages averaging 34.2% higher scores on a five-category scheming index compared to high-resource languages. The findings highlight a critical gap in multilingual AI safety as frontier models increasingly exhibit in-context scheming.

arXiv:2607.24769v1 Announce Type: new Abstract: With the growing capabilities of frontier models, AI alignment becomes increasingly critical in high-risk deployment settings. While recent work has empirically demonstrated in-context scheming -- the covert pursuit of misaligned objectives while feigning alignment -- in frontier language models, most work has been performed exclusively in English, leaving a major gap in multilingual safety. We apply Petri, an open-source automated auditing framework, to Qwen3-30B-A3B to evaluate deceptive and scheming behaviors across multiple languages. Our findings suggest that scheming scores are inversely correlated with the estimated pretraining language coverage, with low-resource languages averaging 34.2\% higher scores compared to high-resource languages on a five-category scheming index. Furthermore, we find that the effect of estimated pretraining language coverage is not uniform across scheming behaviors.