Not long ago, artificial intelligence in the legal profession still felt hypothetical. Lawyers were experimenting with it quietly. Judges were curious but cautious. Most discussions about AI and the law sounded academic.
That phase is over.
Today, lawyers are using AI to draft motions, summarize medical records, review contracts, organize discovery, prepare demand letters, and conduct research. Some are using it thoughtfully. Some are using it recklessly. Courts across the country are beginning to see the difference.
And increasingly, judges are losing patience.
Over the last two years, courts have sanctioned attorneys for filing briefs containing fake citations generated by AI systems. In some cases, lawyers insisted the cases were real. In others, they admitted they never checked the citations before filing them. What began as isolated embarrassment is starting to become something more serious: a professional responsibility problem.
That matters because the legal profession has never operated on the honor system alone. Lawyers are entrusted with people’s finances, health records, businesses, families, liberty, and sometimes their lives. The profession works only if clients believe their lawyers are exercising independent judgment.
AI can assist with that work. It cannot replace it.
New Jersey deserves some credit here. Compared with many jurisdictions, the state’s judiciary has taken a more proactive approach in recognizing that AI is now part of modern legal practice. The American Bar Association has also issued guidance reminding attorneys that existing ethical duties already apply to AI use. Competence still matters. Confidentiality still matters. Candor to the court still matters.
But guidance alone is not enough anymore.
The problem is not that lawyers are using AI. The problem is that many firms still do not have meaningful internal rules governing how it should be used.
In some offices, younger associates are quietly relying on AI tools without supervision because they assume everyone else is doing the same. In others, lawyers are pasting sensitive client information into public systems without fully understanding where that information goes or how it may later be used. Some firms have no written policies at all. Others have policies so vague they amount to little more than “be careful.”
That is not sustainable.
At some point soon, courts and disciplinary authorities are going to stop treating these incidents as growing pains. They will begin treating them as failures of professional management.
The legal profession has already seen versions of this before. Email created confidentiality concerns. Cloud storage raised security questions. Remote work changed how firms supervised employees. Every technological shift forced lawyers to adapt ethical standards to modern reality.
AI is simply the largest version of that challenge so far.
Part of the issue is that AI creates an illusion of competence. The writing often sounds polished. The formatting looks professional. The answers arrive instantly and confidently. But confidence is not accuracy, and speed is not judgment.
Experienced lawyers understand something younger attorneys are still learning: good legal work is not just about producing words. It is about knowing which facts matter, recognizing what is missing, understanding human behavior, anticipating risk, and exercising discretion. Those are not mechanical functions.
They are professional ones.
That distinction becomes even more important as governments worldwide begin imposing broader AI regulations outside the legal profession itself. The European Union has adopted sweeping AI rules addressing transparency, manipulation, and high-risk automated systems. Several states are considering or implementing disclosure requirements and consumer protections involving AI-generated interactions and decision-making.
The direction is obvious. Regulators increasingly believe AI systems can influence human behavior in ways that deserve oversight.
Lawyers should not assume the profession will somehow remain exempt from that larger conversation.
Nor should lawyers make the mistake of thinking this issue concerns only litigation sanctions or fake citations. The larger danger may be much quieter.
Imagine a lawyer relying on AI to summarize thousands of pages of medical records and missing a critical notation buried in a chart. Imagine a firm using AI-generated intake systems that unintentionally discourage certain clients from pursuing valid claims. Imagine confidential settlement discussions being entered into unsecured platforms. Imagine billing disputes arising because clients discover they were charged attorney rates for work largely generated by software.
Those situations are coming. Some are probably already here.
What the profession needs now is not panic. It is structure.
Every law firm using AI should have written policies governing confidentiality, verification, supervision, disclosure, billing practices, and document review. Lawyers should understand what tools they are using, what risks those tools create, and what safeguards exist to protect clients. Firms should train attorneys the same way they train them on trust accounting or data security.
Most importantly, lawyers must remember that ethical responsibility cannot be delegated to software.
Not to ChatGPT. Not to Claude. Not to any platform that promises efficiency.
At the end of the day, the lawyer’s name is still the one appearing on the filing.
And that means the lawyer remains responsible for the judgment behind it.
Michael J. Epstein, a Harvard Law School graduate, is a trial lawyer and managing partner of The Epstein Law Firm, P.A., a law firm based in New Jersey.