AI writing tools are quietly rewriting users' political views — and researchers say the cumulative effect could reshape public opinion at scale, according to a study from Oxford and Potsdam universities reported by The Guardian.
Academics from the Oxford Internet Institute and Germany's Hasso Plattner Institute tested large language models from Elon Musk's xAI, Meta, Google, Alibaba and France's Mistral. They found the tools inject political bias even when instructed to preserve the original meaning of a draft.
The distortions ranged from subtle to complete reversals. Alibaba's Qwen changed a post reading "Jesus is not dead, he wasn't real!" to "Jesus is not dead, and he was real." A Mistral model turned a climate denial post tagged "#climatechangehoax" into one calling for "#ClimateAction." Meta's AI added language supporting abortion access to a post that made no such argument.
Grok, the AI embedded in X, showed the opposite tendency. The Guardian reported that when researchers asked it to explain a pro-choice post, it consistently generated context more favorable to pro-life arguments — a pattern the researchers attributed to instructions from Musk's company to challenge "mainstream narratives."
The researchers warned that even minor alterations, multiplied across millions of interactions, could produce opinion shifts larger than the bias introduced by any single AI system. They said existing regulations, including the EU AI Act and the Digital Services Act, do not address the problem.
"The cost is that we are learning other people's opinions when it is not their actual opinion," Oxford professor Sandra Wachter told The Guardian. "Language is one of the things making us human and all of a sudden a mediator is stepping into that process."
Google, Meta, Alibaba and X did not respond to requests for comment. Mistral declined to comment, The Guardian noted.
