Thuringia’s proposed AI police powers are broad enough to worry the police union

Police unions rarely object to receiving more operational powers. That is what makes the response to Thuringia’s proposed police law notable. The state branch of the GdP warned of serious fundamental-rights risks and called for firm constraints on AI-assisted surveillance, according to netzpolitik.org.

The problem with broad authorization is not only what today’s system can do. Legal language that permits automated analysis can become a standing permission slip for tomorrow’s models, datasets and forms of inference. Combining records can reveal associations, routines and predicted behavior that no individual database contains.

A durable law would specify purpose, data sources, retention, human responsibility, logging, audit access and the threshold for deployment. It would also distinguish tools that organize known evidence from systems that generate suspicion or classify people. Treating both as generic “AI” avoids the hard policy choices.

The important boundary is not AI versus no AI. It is assistance versus suspicion at scale.

Vague powers create two failures at once: citizens cannot predict how they will be examined, and police officers cannot know where legitimate use ends. If lawmakers cannot describe the allowed workflow precisely enough to audit it, the technology should not receive blanket statutory authorization.