Our Firm

AI Will Make Lawyers More Accurate, Not Less

When a self-driving car is involved in an accident, it makes the news. Regulators call hearings, and the public wonders whether we have made a terrible mistake.

Yet a growing body of research indicates that self-driving cars are involved in fewer accidents than human drivers. Human drivers cause crashes every day, and almost none of them make the news. We have simply accepted human error as the cost of getting around.

The legal profession is approaching its own version of this moment. When a lawyer files a brief with a fabricated AI citation, it becomes a cautionary tale. Courts issue sanctions, and legislatures take notice.

But before AI, lawyers made mistakes too. We miscited cases. We cited authority that did not quite support the proposition, misstated facts, and mischaracterized documents. And we missed the important sentence on page 4,374 of a production. No lawyer I know is immune, and those errors never made headlines because human error is the baseline we have always lived with.

In the hands of a trained lawyer, AI changes that baseline for the better. It can cross-check a brief against the record, surface the one transaction in thousands of pages of financial records that deserves a closer look, and catch the inconsistency a tired attorney would miss at 11 p.m. None of that replaces judgment, but it leaves more time for it. In my view, the result is work that is more accurate and more thorough than what we produced before.

I am talking about lawyers using AI, not non-lawyers using it to generate legal arguments they do not understand. A trained attorney knows what a correct answer should look like and can catch the output that is wrong. Someone without that training cannot tell a sound citation from a fabricated one, and those filings create more work and more errors for everyone. In those cases,
what is missing is the expertise to supervise the technology.

Supervision also means knowing the tool’s weak spots. Lawyers using AI need to stay current on where it fails, including fabricated citations, misquoted statutes, overstated holdings, and invented facts when the record is thin. Those limits shift as the tools improve, and every citation and factual assertion still gets checked against the source.

There is growing focus, and new legislation, on disclosing AI use to clients and to the court. I support that. Regulating AI is really a matter of professional responsibility: competence, candor to the court, and supervision of our work. This article focuses on accuracy because it is one part of that larger picture. As for disclosure, I used AI to help me draft the first version of this article.
I dictated the argument and the points I wanted to make, and the AI helped me put them on the page.

But we should measure AI against the right benchmark: whether lawyers using it make fewer mistakes than lawyers working without it, not whether it ever errs. Regulating against an imaginary standard of perfection risks discouraging tools that could make the profession more reliable.

Hold AI and lawyers using it to a high standard. But let’s also be honest about the standard that
came before it.



 Back to All Insights