The legal industry is not immune to the influence of artificial intelligence. A US lawyer and legal blogger believes that Claude AI can adjudicate complex legal cases and theoretically even serve as a Supreme Court Justice.

Adam Unikowsky, an acclaimed lawyer who was included in Lawdragon's "Top 500 Litigators in America" list for 2024, made this bold statement in his legal newsletter on Substack.

"Claude is now fully capable of serving as a Supreme Court Justice. When used as a legal assistant, Claude's insights and accuracy easily rival those of human assistants, while surpassing them in efficiency," he wrote.

He arrived at this conclusion by downloading legal summaries of all Supreme Court cases during his term and inputting them into Claude 3 Opus, followed by posing several subsequent questions.

This AI tool comes from Anthropic, a company that released the Claude 3 model family in March. At the time, this model surpassed previous benchmarks. However, the company's release of Claude 3.5 Sonnet once again exceeded expectations.

How legal experts evaluate Claude AI's legal capabilities

Firstly, the AI tool was tasked with adjudicating cases, and the responses consistently demonstrated its ability to make correct rulings.

"When it 'errs'—that is, when its rulings differ from those of the Supreme Court—its rulings are always reasonable."

In the analysis of 37 substantive cases, the AI was found to have made the same rulings as the Supreme Court in 27 cases. Unikowsky stated that in the remaining 10 cases where it made incorrect rulings, he was often "persuaded by Claude's analysis."

After using the tool for more work and requesting its rulings, the lawyer found that Claude "clearly has the ability to adjudicate complex cases."

He even went further to place the answers from the chatbot assistant "at or above the level of a human Supreme Court assistant."

"Claude not only provides reasonable suggestions and drafts judicial opinions, but also effortlessly generates new legal standards and identifies methodological errors in expert testimony."

Unikowsky believes that "Claude occasionally makes mistakes, but so do humans."