Skip to content
IFAI

Industry

Supreme Court chief justice weighs in on AI in the courtroom

Published Reach 50

Chief Justice John Roberts just warned that artificial intelligence requires "caution and humility" across the federal judiciary. While acknowledging AI can broaden access to justice and speed up research, he stressed that hallucination risks mean human discretion and judges aren't going anywhere anytime soon.

Supreme Court chief justice weighs in on AI in the courtroom
Image: siliconangle.com

Even today, counsel tables inside the Supreme Court still feature traditional white goose quill pens. The inkwells disappeared decades ago, but the quills remain as historic souvenirs. In a legal system famously allergic to change, technology usually moves in slow motion—until artificial intelligence arrived.

Chief Justice John Roberts released his 13-page Year-End Report on the Federal Judiciary, devoting the centerpiece of his annual address to AI. Roberts urged the legal profession to approach the technology with "caution and humility." His headline takeaway: human judges are not going obsolete anytime soon, but everyday judicial practice is about to be completely overhauled.

From touch-tone case lookups to hallucinated briefs

The federal court system has seen plenty of awkward tech rollouts. Roberts pointed out that judges' chambers did not get personal computers until 1989. That same year, the courts debuted the Voice Case Information System (VCIS), letting the public dial in on touch-tone phones to listen to synthesized case updates.

Fast forward to today: generative models are acing the bar exam and analyzing thousands of case files in seconds. The catch, of course, is hallucination.

AI programs have a nasty habit of making up nonexistent precedents out of thin air. Just last week, papers tied to former Trump lawyer Michael Cohen were revealed to contain fake citations produced by an AI tool. Incidents like this made headlines all year, forcing lower courts to scramble for guardrails.

Some are already taking strict measures. Last month, the 5th U.S. Circuit Court of Appeals in New Orleans unveiled a proposed rule requiring lawyers to certify that they either drafted their briefs without AI or had humans manually verify the accuracy of any machine-generated text.

Why judgment beats raw compute

Roberts characterized AI as a distinct "mixed blessing." On one hand, it holds immense promise for expanding access to justice, lowering costs for litigants who cannot afford high-priced counsel, and speeding up legal research.

On the other hand, the law demands something machines simply do not possess: human discretion.

Judicial decisions rely on empathy, moral nuance, and real-world context that algorithms cannot emulate. While Roberts predicted that "human judges will be around for a while," he made it clear that trial-level courts will bear the immediate brunt of this technological transition.

Why it matters

Roberts' intervention provides a necessary reality check for the legal tech boom. Instead of indulging in sci-fi tropes about automated justice, he draws a sharp boundary around legal accountability.

For everyday citizens and solo practitioners, AI remains a powerful equalizer. It democratizes legal research, giving small teams capabilities that used to require a small army of paralegals. But in a courtroom, trust is everything. A single fabricated citation can wreck a lawyer's credibility or derail a case entirely.

The 5th Circuit's rule is likely a blueprint for what is coming everywhere else: you can use whatever software you want, but a human must sign on the dotted line and take personal responsibility for the facts.

We think Roberts struck a remarkably pragmatic tone here. By rejecting both mindless tech evangelism and outright Luddism, he sent a clear signal to the legal world: AI can do the heavy lifting in research, but the ultimate burden of judgment still rests entirely on human shoulders.

Sources (2)

  1. [1]
    supremecourt.gov(opens in a new tab)Originalsupremecourt.gov
  2. [2]

Best of AI, daily