Why Are Lawyers Still Getting Sanctioned for AI Hallucinations? It's Not the Technology
Lazy lawyers were around before AI
5/25/20265 min read
Why Are Lawyers Still Getting Sanctioned for AI Hallucinations? It's Not the Technology
By Angeli Fitch, AI Compliance & Ethics Attorney | 20+ Years of Trial Experience | Creator of the State Bar-Approved CLE Course "AI Ethics for Attorneys"
What is an AI hallucination, and why does it keep showing up in court filings?
An AI hallucination is when a generative AI tool produces content that sounds authoritative and is factually wrong. In legal practice, the most damaging version of this is fabricated case citations — AI-generated references to court opinions that do not exist, attributed to real courts and real judges, formatted to look exactly like legitimate legal authority.
The numbers are not hypothetical. Stanford researchers found hallucination rates between 69% and 88% across leading AI models on legal tasks. More than 486 cases worldwide have featured AI-generated hallucinations in court filings. Over 324 of those are in U.S. courts.
In September 2025, a Los Angeles attorney received a $10,000 sanction after 21 of 23 case citations in an appellate brief were fabricated by ChatGPT. An Arizona lawyer was ordered to personally notify three federal judges that she had falsely attributed opinions to them. A Colorado attorney lost his license for 90 days — and sent a text message to his paralegal acknowledging he had not checked the AI's work before filing.
These are not isolated incidents. They are a pattern. And the pattern has a cause that has nothing to do with the technology.
Is this an AI problem or a lawyer competence problem?
It is a lawyer competence problem that AI has made impossible to ignore.
Every ethical duty these attorneys violated existed before ChatGPT launched in November 2022. ABA Model Rule 1.1 has always required lawyers to exercise the legal knowledge, thoroughness, and preparation reasonably necessary for competent representation. The obligation to verify case citations before filing them is not a new requirement created in response to AI. It is foundational to the practice of law — and has been for over a century.
Lawyers still Shepardize. That term remains standard practice in 2025, referring to the process of validating citations through Shepard's Citations on Lexis, KeyCite on Westlaw, or BCite on Bloomberg Law. These citator services confirm whether a case is still good law. Law schools teach this as a basic skill. It has always been the attorney's responsibility to perform it — regardless of where the initial research came from.
ChatGPT itself warns users to verify its output. The ABA's Formal Opinion 512 states plainly that lawyers using generative AI must understand its benefits and risks and remain personally responsible for all AI-assisted work. That opinion did not create a new standard. It restated the existing one as applied to a new tool.
What excuses have judges heard from lawyers sanctioned for AI errors?
The excuses appearing in disciplinary records and sanctions orders fall into a consistent set:
"I didn't know AI could fabricate cases." This is difficult to accept as a defense in 2025. The risk of AI hallucination in legal research has been the subject of bar ethics opinions, CLE programs, legal publication coverage, and law school curricula for years. Ignorance of a widely publicized risk is not a mitigating factor — it is the competence failure.
"I trusted the technology." Trust is not verification. Trusting a research tool — any research tool — without checking its output has never been a defense to a competence violation. The standard is the same whether the unverified source is an AI, a paralegal's memo, or a citation from opposing counsel's brief.
"I was under time pressure." Every lawyer working under deadline pressure has faced the choice between thoroughness and speed. Filing court documents without reading them resolves that tension the wrong way — and it does so at the client's expense.
"My paralegal did it." Under Rules 5.1 and 5.3, attorneys are responsible for supervising both other lawyers and nonlawyer staff. Delegating research to a paralegal who uses AI does not transfer the duty of competence. It extends the supervisory obligation.
None of these would have worked before AI existed. They do not work now.
What does this mean for legal malpractice exposure?
Malpractice carriers are paying attention to this in ways that should concern practitioners. Firms without documented AI verification protocols face heightened scrutiny at renewal, potential premium increases, and in some cases exclusions for AI-related claims.
The malpractice theory is straightforward. A client harmed by an AI hallucination that an attorney filed without verification has a viable claim for breach of the duty of care. The harm is not speculative — sanctions, adverse rulings, damaged credibility with the court, and prejudice to the client's position are all concrete and compensable.
The argument that AI use somehow lowers the competence standard has been rejected by every state bar that has addressed it. Several bars have reached the opposite conclusion: AI use raises the competence standard, because attorneys must now demonstrate not just legal skill but sufficient understanding of their tools to deploy them responsibly.
What does competent AI use actually require from a lawyer?
At minimum, competent use of AI in legal practice requires:
Understanding how the tool works well enough to know where it fails. AI language models are not search engines. They do not retrieve documents — they generate text that resembles documents. An attorney who does not understand this distinction cannot reliably evaluate AI output.
Direct human review of all AI-generated content before it goes anywhere — to a client, to opposing counsel, or to a court. This is not optional and cannot be delegated away.
Independent verification of every legal citation through a citator service, regardless of how confident the AI output appears. Hallucinated citations are formatted correctly. They look real. The only way to know they are not is to check.
Supervision of every staff member using AI on client matters. If your paralegal is using AI to draft or research, your supervisory obligations extend to that use.
Documentation of the verification steps taken. When a disciplinary body or malpractice carrier asks what you did to verify AI-generated content, "I trusted the output" is not an answer.
What is the broader lesson for the legal profession?
AI is a mirror. What it is reflecting back to the profession is not a new problem — it is an existing one that faster, more confident-sounding technology has made impossible to hide.
Over-delegation without supervision. Failure to read work product before filing it. Treating thorough preparation as optional under deadline pressure. These habits existed before generative AI. AI has simply removed the cover that less capable tools once provided.
The law firms and solo practitioners who will thrive in the AI era are not necessarily the ones who adopted the technology earliest. They are the ones whose underlying commitment to competent, ethical practice was strong enough to hold when the technology introduced new ways to cut corners.
That commitment is not a constraint on using AI effectively. It is the precondition for it.
What should lawyers do right now to protect themselves and their clients?
Establish a written AI use protocol for your practice before the next filing deadline, not after a problem surfaces. That protocol should address which tools are approved, what data can be input, what verification steps are required, and who is responsible for each step.
If you teach, supervise, or manage other attorneys, make sure they understand the same. The sanctions in the headlines were not all solo practitioners making individual errors. Some were partners whose supervision failures extended liability to the firms they led.
If you want structured guidance on building an AI compliance framework that maps to your existing ethical obligations, that is exactly what I address in my State Bar-approved CLE course, AI Ethics for Attorneys. [Contact me] to bring this training to your firm or organization.
Angeli Fitch is an AI Compliance & Ethics Attorney and trial lawyer with more than 20 years of experience. She is the creator of the California State Bar-approved CLE course "AI Ethics for Attorneys" and advises law firms and legal professionals on ethical AI adoption, compliance, and governance. She is Of Counsel at Infinity Law Group and available for speaking engagements, CLE instruction, and advisory work.