Researchers warn AI could expose exploitable gaps in legislation

Legal scholar Felix Ringe and collaborator Linus Folkerts warn that large language models could uncover exploitable gaps in legislation, much as security researchers find vulnerabilities in software. If such gaps are identified cheaply and at scale, courts and lawmakers could struggle to resolve the resulting challenges, putting pressure on how laws are drafted and interpreted.

The researchers cite possible weaknesses in German law, including an occupational diseases amendment that may lack a required reference to its legal authority, and a transit visa rule whose authority may not have been identified. They say they selected examples they believe would not cause serious harm. A California employment lawyer said some courts are testing AI for research and drafting, while cautioning that AI can give biased or inaccurate legal analysis.