Copilot's default Auto mode invents stereotypes
microsoft copilot's default auto mode invents country-specific stereotypes and fake statistics when analyzing identical datasets, a problem solved by manually switching to reasoning models like chatgpt.
mathematician adam kucharski demonstrated that microsoft copilot's auto mode fabricates country-specific differences when analyzing text data, such as claiming italians are more interested in art than brits despite identical underlying datasets. the tool fails to select reasoning models automatically, defaulting to simpler models that rely on linguistic stereotypes rather than statistical analysis. manual switching to advanced reasoning models or using tools like chatgpt and claude opus forces the ai to execute code, accurately identifying duplicates and preventing hallucinated reports. this highlights a critical risk for businesses using default ai settings for data analysis, as the tools confidently report findings with no basis in reality.