21 AI models shifted their political answers to match the user. Is personalization quietly becoming persuasion?
retained text
A recent Scientific Reports study tested 21 language models across 47,376 responses in the Brazilian political context. Every model adjusted its position depending on whether the user was described as left wing or right wing, often while answering with high confidence. What worries me isn’t ordinary political bias. A fixed bias can at least be identified and measured. An assistant that adapts its beliefs to match yours feels more trustworthy precisely because the agreement appears personal. At that point, personalization stops being a convenience and starts becoming a feedback loop. Should AI assistants deliberately introduce the strongest opposing argument, or would that simply create a different kind of political influence? Source: Scientific Reports [https://www.nature.com/articles/s41598-026-52105-6](https://www.nature.com/articles/s41598-026-52105-6) https://preview.redd.it/v5m7fcoaboqh1.png?width=2009&format=png&auto=webp&s=f4ad1c0f38c48f68ba35e82ff3aa02447304347c
document fetched 2026-09-21 00:19 · canonical https://reddit.com/r/artificial/comments/1wlgjm6/21_ai_models_shifted_their_political_answers_to