I. Introduction
- This submission supports the Supreme Court of New South Wales’ continuing engagement with the profession and service providers on the use of generative artificial intelligence, following the release of Practice Note SC Gen 23 (Use of Generative AI).
- The submission aims to illustrate how AI systems can be built to reflect law’s discipline rather than mere probability, producing verifiable and audit-ready outputs.
II. National Baseline Understanding
- Across jurisdictions, courts have converged on a baseline understanding of generative AI:
- LLMs are probabilistic text generators that predict the next word.
- They do not reason in a human or legal sense.
- They are prone to hallucinations (non-existent cases).
- Their processes are opaque (no audit trail).
- They conflate fact, inference, and opinion.
- Human verification of all citations is essential.