Core judicial functions like interpreting precedent, legal analysis, and decision-making, must not be delegated to GenAI. Judges are responsible for their work and decisions. This entails verifying every citation, ensuring the accuracy of factual assertions, and reviewing all conclusions. This is a fundamental rule of law principle that underpins judicial independence.

Judges interpret law, exercise discretion, weigh credibility, assess context, engage in reasoning, and exercise empathy. GenAI can process information and identify patterns but cannot and should not be used to engage in core judicial functions. GenAI is not ‘responsible’ for its output. 

GenAI-generated text can sound confident even when it is wrong. There may be a tendency to over-rely on computer-generated output because it appears objective and authoritative. Treat GenAI output like a draft from an intern or new law clerk – a potentially useful starting point but one that requires thorough review and verification.

Fairness is another fundamental rule of law principle. AI systems can intentionally or inadvertently introduce biases that disproportionately impact certain groups. For example, there are reported cases of real estate companies using AI algorithms that were alleged to have had a disproportionate impact on the ability of minority groups to gain access to rental housing. The algorithms were based on credit data that reflected historical disparities in the ability to obtain mortgages. There are many different kinds of bias that can impact the fairness of an AI system, including sampling bias (non-representative training data), statistical or mathematical bias (systemic non-random errors in model predictions), discriminatory bias (the reproduction of human prejudices in training data or algorithms), and the user’s cognitive bias (e.g., automation and confirmation bias). 

GenAI cannot reason through unsettled law or navigate genuinely novel or nuanced legal issues the way a human can. The use of GenAI for such questions raises the risk of delegating core judicial functions as well as importing errors that are difficult to detect because there is no established correct answer to verify against.

Judges are responsible for ensuring appropriate GenAI use by chambers staff. This requires clear policies and communication, training, and active oversight. Staff must understand that the judge is accountable for all final work product.

GenAI can be helpful for suggesting language alternatives, editing for tone and clarity, summarizing documents, organizing facts into timelines, and in some instances, drafting text. GenAI is adept at these types of tasks, but it can and does make mistakes. For that reason, all GenAI output requires human review and judgment to check and verify its output.