Supreme Court justices are not currently using artificial intelligence in their work due to security concerns, though members of the court have expressed growing interest in the technology. During congressional testimony in July 2026, Justice Amy Coney Barrett stated that while the court is not yet utilizing AI to gain efficiencies, it is something they are studying. Justice Elena Kagan noted that the court is closely examining what future rules and best practices should be for how justices, clerks, and assistants might appropriately use AI to ensure it is used appropriately without creating new dangers.
The court's internal caution contrasts with the increasing use of AI by attorneys and analysts to predict judicial outcomes. A study released in July 2026 by researchers Hayley Stillwell and Sean Harrington found that while AI models like ChatGPT-5 and Claude Sonnet 4.5 can identify broad controversies, they frequently struggle with the specific legal questions that determine rulings. The study also found that AI tools systematically overpredict ideological division, often anticipating 6-3 decisions even though approximately 43% of the court’s merits decisions are unanimous.
Individual justices have addressed AI through both formal comments and humor. In March 2026, during oral arguments for Jules v. Andre Balazs Properties, Justice Samuel Alito asked attorney Adam Unikowsky if the court should use the chatbot Claude to decide the case; Unikowsky declined the suggestion. Justice Sonia Sotomayor, speaking at the University of Alabama in April 2026, suggested that if AI can perfectly predict outcomes, it might indicate the court is too predictable and not sufficiently "opening our minds to new ideas."
The scale of AI's potential influence is reflected in research identifying the specific writing styles of the justices. A study in the Journal of Legal Analysis reported that deep learning models can predict the authorship of unsigned "per curiam" opinions with 91% accuracy, drawing from a database of over 4,000 opinions written between 1994 and 2024. This capability potentially removes the anonymity of unsigned rulings, which comprise 10% to 15% of the court's output. For the public, this means that even when the court attempts to speak as a single, anonymous voice, technology can now attribute specific sections to individual justices, potentially changing how legal strategies are tailored and how judicial influence is measured.
The concrete day-to-day impact is currently most visible in the trial and appellate levels, where some jurisdictions, such as New York, have implemented policies prohibiting judges and staff from entering confidential information into public generative AI models. As the Supreme Court continues its study, future policy considerations could involve style guides or rules for how law clerks utilize automated tools. While no specific deadline for a court-wide AI policy has been set, the justices have indicated that the ongoing study focuses on balancing efficiency with the need to prevent dangers, such as fabricated facts or citations; a separate case involving a government report containing fake legal references resulted in a $97,000 refund.