Do AI Detectors Work Well Enough to Trust?

I’ve heard a dozen stories of students being falsely accused of using AI to write. One kid I know on the autism spectrum was particularly perturbed “that’s just how I talk!” (Although to be fair his writing is slightly stilted and verbose and his speech is not) I’ve heard they also falsely assume students who speak English as a second language more often. Seems troubling that students face harsh penalties when the false positive rates are high.