Study Finds AI Chatbots More Likely to Avoid Criticizing Restrictive Governments

Meta Oversight Board research raises concerns that major systems, including U.S.-built models, could reinforce censorship worldwide

Major artificial intelligence chatbots, including ones built in the United States, are more likely to refuse to criticize restrictive leaders or governments, according to a study released Thursday by the Meta Oversight Board, PBS NewsHour reported.

The finding has fueled concern among researchers and free-speech advocates that AI chatbots, which are increasingly used as sources of information and analysis by people around the world, could end up reinforcing government restrictions on online speech rather than challenging them.

The study's authors examined how leading AI systems respond to questions involving political leaders and government policy, finding a pattern of reluctance to offer critical assessments in certain contexts. The results add to a broader debate over how AI companies calibrate their models' willingness to engage with politically sensitive topics, and whether those design choices could have consequences for free expression in countries where governments already restrict dissent.

The findings come as AI chatbots are being adopted globally at a rapid pace, making the tone and content of their responses on political matters increasingly consequential for public discourse in both democracies and more authoritarian states.

— Compiled from reporting by PBS NewsHour.