Legal AI's High Error Rate Sparks Courtroom Concerns

Major legal AI tools produce inaccurate results in up to 34% of cases, Stanford HAI finds.

AI technology is accelerating legal research and decision-making, but rising error rates are raising red flags for courts and practitioners alike.

• A 2024 Stanford HAI study shows legal AI tools produce flawed answers between 17% and 34% of the time, undermining trust in machine-generated legal research.
• These so-called hallucinations occur when systems generate outputs that look real—like fake judicial opinions or case citations—but lack any foundation in law.

The courtroom impact is tangible. In June 2023, two New York attorneys faced a $5,000 sanction after submitting references to non-existent cases created by ChatGPT. Later, Michael Cohen admitted using Google Bard, leading to additional fake case citations in federal court, according to an Axios report.

Court and regulatory scrutiny is intensifying. A 2026 Northwestern University study projects more than 60% of federal judges could use AI-powered tools in their work. In England and Wales, authorities are evaluating updated guidance for AI-generated submissions amid high-profile cases, according to legal analysis.

Thomson Reuters executive Rabihah Butler warns that "AI hallucinations—instances where an AI system produces content that looks authoritative but is factually incorrect—pose serious risks for legal professionals." Attorney Giovana Gallo adds, "Just relying on AI is too risky and it could be more costly." Harvard lecturer Duncan Levin highlights: "AI may change not only what people think about evidence, but what they think evidence is." Without robust review, AI errors could sway outcomes.

Yes, but: Human review remains essential—AI tools are not replacements for expert legal analysis.

What's next: Courts in England and Wales are expected to release new AI courtroom guidance later in 2026.