A new preprint on arXiv asks a pointed question: can artificial intelligence agents make law? The study, titled "Can AI agents make law? Evaluating AI legal relevance and creativity," focuses on large language models—the systems behind modern chatbots—and whether they can build on their legal training to draft rules for situations that existing text never anticipated.
The abstract notes that these models are trained on vast amounts of legal text, but the core inquiry is whether they can go beyond that material. The research appears to evaluate both legal relevance and creativity, suggesting the authors are testing not just whether AI outputs are legally sound, but whether they can generate novel legal constructs.
Because the abstract is truncated, specific methods and results are not available. The study's significance lies in its framing: it probes a boundary of AI capability, asking whether machines can contribute to lawmaking in contexts where precedent and existing text fall short.