Researchers at Apple discovered that LLMs fail to update probabilistic beliefs consistently when presented with new evidence. The team measured this using a novel information processing gap to quantify deviations from Bayes' rule. These internal inconsistencies hinder rational decision-making in high-stakes fields.