Google DeepMind and UPenn build AI text detectors that show their work

AI text detectors already help decide who gets into a school, whose paper survives peer review, and which posts stay online. Most of them cannot explain a single decision they make. A research team from Google DeepMind and the University of Pennsylvania wants to change that. Their new method turns an opaque detector into one that reasons in concepts a human can read, and the accuracy cost is surprisingly small. The paper, titled “Building Auditable AI Text Detectors with Sparse Autoencoders,” was submitted on October 21, 2026. It has been accepted at EMNLP 2026, one of the major venues for natural language processing research. Opening the black box The problem the researchers describe is a familiar trade-off. The best-performing detectors are fine-tuned neural classifiers, and their internal logic is essentially sealed. The alternatives are transparent tools built on stylometric and linguistic features, such as patterns in word choice and sentence structure. Those you can inspect. According to the paper, they also trail the neural models by wide margins. The team’s answer is a pipeline that keeps the neural model’s knowledge but translates it. The process starts with a fine-tuned d...