[Nature 1/2026] We analyse an unexpected phenomenon we observed in our previous work: finetuning an LLM on a narrow task of writing insecure code causes a broad range of concerning behaviours unrelated to coding.

LLMs transmit traits to other models via hidden signals in data. Datasets consisting only of 3-digit numbers can transmit a love for owls, or evil tendencies.

Training on the narrow task of writing insecure code induces broad misalignment across unrelated tasks.

We propose a benchmark to measure whether a language model is truthful in generating answers to questions.