Not unlike a high-powered Lamborghini operated by a new motorist without a seatbelt or owner’s manual, recent cases illustrate a common theme in an avalanche of AI legal research gone wrong: the technology isn’t usually the issue—it’s how it’s used.