Lawyer Disbarred for Using AI: What Actually Happened
Mata v. Avianca: what actually happened when lawyers filed AI-invented cases, the sanction imposed, and the duties that applied.
Last Updated: August 2026
Legal hallucinations are incorrect or fabricated citations, summaries, or analyses produced by generative AI tools when they lack sufficient training data or context for a specific task. In the legal sector, these errors often stem from users failing to verify output before submission, resulting in severe professional consequences. The risk is not inherent to technology, but to its application without a verification framework. AiBuildrs provides the infrastructure required to prevent these failures by grounding language models in proprietary business data rather than open web models. By combining human-in-the-loop workflows with custom indexing, these systems ensure that every output adheres to established standards. The team has completed over 200 successful AI implementations across diverse industries, including professional services and healthcare. These deployments focus on replacing manual oversight with hard-coded logic and RAG, which stands for retrieval-augmented generation. Grounding a system in your own archives and requiring it to cite them narrows the room for invention. It does not remove the need for a human to check what is cited.
Key Takeaways
- Inaccurate reporting: No attorney was disbarred in Mata v. Avianca. The court imposed a $5,000 penalty.
- The risk of generic tools: General models invent case law that reads correctly, which is precisely what makes it dangerous.
- Verification remains mandatory: The duty to check what you file predates the technology and is unaffected by it.
- Name the checker: A policy that does not say who verifies before filing is not a control.
- The failure was process, not software: Six citations were filed that nobody opened.
What was the Mata v. Avianca AI case?
The opinion (2023) records six fabricated decisions cited in a filing, each attributed to a real judge, none of which existed.
The Mata v. Avianca case involved attorneys for plaintiff Roberto Mata who submitted a legal brief containing multiple fictitious case citations generated by ChatGPT.
The presiding judge ultimately issued a $5,000 sanction against the attorneys, including Steven A. Schwartz, for failing to verify the authenticity of the information provided by the tool.
The incident gained significant attention because it highlighted the risks of using generative AI for specialized research without human oversight. The attorneys had asked the platform to locate prior court decisions, and it responded with names and summaries of cases that did not exist. When the defense team alerted the court that they could not find these precedents, the plaintiffs initially doubled down on their validity.
This failure to perform basic due diligence resulted in public censure and a financial penalty for the firm. It serves as a clear warning regarding the limits of current large language models in professional settings.
Was the lawyer actually disbarred?
No attorney was disbarred. In Mata v. Avianca (2023), the court imposed a $5,000 penalty, holding it "sufficient but not more than necessary to advance the goals of specific and general deterrence", and required the lawyers to notify the judges named as authors of the fabricated opinions. Professional misconduct can range from honest errors to severe ethics violations. Disciplinary bodies assess the intent and the impact of the actions taken by the practitioner.
Lesser penalties like reprimands or fines serve as a formal correction for behavior that falls below the expected standard of practice. This distinction is vital for understanding how the legal system responds when AI tools introduce new risks into professional work.
Why did ChatGPT create fake legal cases?
Large Language Models (LLMs) like ChatGPT from OpenAI function by calculating the statistical probability of language to predict the next plausible word in a sequence. These systems are not designed to search a database or verify facts against a source of truth.
When the model lacks a specific reference for a legal case, its focus on textual fluency forces it to generate a response that mimics the appearance of a real citation. This technical process often leads to AI hallucinations, where the system produces highly convincing but entirely fictitious content.
Because these tools prioritize linguistic structure over accuracy, they can invent case names, judge opinions, and statutes that look correct to a casual observer. This occurs because the model prioritizes the coherence of the prose rather than the validity of the data it retrieves. Relying on an LLM for factual verification in professional work without human oversight poses a significant risk to data integrity.
What are the ethical duties for lawyers using AI?
AiBuildrs treats verification as a workflow step rather than a policy line, because a policy nobody executes is what produced this case.
Lawyers maintain their duty of competence and duty of diligence when using AI in legal practice. These core obligations require attorneys to supervise the output of any technology to ensure it is accurate and valid.
The ABA Model Rules dictate that an attorney is always responsible for the work product they submit to a court. The presence of a machine does not shift this burden or excuse mistakes.
Rule 1.1 of the ABA Model Rules of Professional Conduct requires that a lawyer provide competent representation. This duty includes staying informed about the benefits and risks of relevant technology. If a lawyer uses AI to research or draft documents, they must have the expertise to recognize when the tool is producing errors.
Rule 5.1 and Rule 5.3 further mandate that lawyers supervise their staff and non-lawyer assistants. These rules apply equally to automated software.
Verifying the accuracy of any AI-generated work product is a fundamental aspect of this supervision. A model may cite non-existent cases or misapply legal standards because it predicts text rather than understands law. You must validate all citations and arguments against reliable primary sources.
An AI is a tool, not a licensed professional. Your signature on a filing represents your personal guarantee that the content is true and grounded in law.
Relying on an unverified output is a breach of your ethical duties and invites significant professional risk. Professional services firms that prioritize precision use custom systems to limit these liabilities.
How can law firms use AI safely?
Every AiBuildrs build for a regulated client names the person who checks the output before it leaves the building.
Law firms safely use AI by establishing a formal AI governance policy that mandates human oversight for every output. You should avoid public models that risk data leaks and instead use bespoke systems built on your own private, verified data.
A firm must define clear rules for how tools handle client materials to protect confidentiality. (2024), adopting a risk-based framework is the most effective way to secure legal workflows while maintaining compliance.
Applying this requires specific controls. For example, your staff can use AI to generate deposition summaries from case transcripts, provided a qualified attorney reviews the output before it enters a filing.
Bespoke systems function within your own environment, ensuring that sensitive information never leaves your control to train a public model like those from OpenAI or Anthropic. This approach ensures that performance remains consistent while closing gaps in data security.
What's the real risk of not using AI in law?
The biggest risk of avoiding AI in law is falling behind competitors who use it to improve efficiency, reduce costs, and deliver faster client outcomes.
Firms that embrace this change focus on automating repetitive tasks like contract review and document discovery. This allows partners to reallocate high-value attorney time to complex strategy and client guidance.
If your competitors offer lower fees because they use Bespoke AI Systems to cut overhead, your pricing model will eventually lose its appeal. McKinsey & Company (2023) notes that AI helps firms capture significant value by augmenting professional work. Staying static is not a neutral choice in a market that rewards speed and precision.
Frequently Asked Questions
Is it illegal for lawyers to use AI?
No, it is not illegal for attorneys to use AI tools in their legal practice. Lawyers must remain competent regarding technology under their professional conduct rules. The legal risk arises when AI outputs are used without human review or verification.
Misrepresenting AI-generated work as human-verified analysis violates professional standards. Firms should focus on building oversight into their internal workflows to maintain compliance while increasing efficiency.
What is the most common reason for an attorney to be disbarred?
The most common cause for disbarment involves client neglect, misappropriation of funds, or dishonesty in legal filings. While AI is a new variable, the ethical duty to provide competent representation remains the same.
Attorneys face disciplinary action when they fail to verify the accuracy of information submitted to a court. Reliance on non-verified technology does not exempt a practitioner from the responsibility to ensure all filings are accurate and professional.
Can lawyers use ChatGPT?
Yes, and many do. Bar guidance across jurisdictions treats it as a tool a lawyer remains responsible for, not a delegate. The duty that matters predates the technology: you verify what you file. In Mata v. Avianca (2023), the failure was not using a model; it was filing six citations nobody checked against a real reporter before signing.
What specific AI tools are safe for legal research?
Safety here is a workflow property, not a product one. A research tool that shows its sources and links to the actual reporter lets a lawyer verify in seconds; one that composes an answer without them requires the same verification with more effort. Whichever you use, the checkable test is whether every authority in the draft can be opened and read before it is filed.
How do I create an AI usage policy for my law firm?
A sound AI usage policy begins by defining which tools are permitted for client work and which are forbidden. The policy must mandate human review for every output used in legal proceedings.
It should include clear instructions on handling client metadata and anonymizing documents before feeding them into any model. Your policy should clarify that the attorney remains fully responsible for the accuracy of all work submitted to the court.
What is the difference between an AI hallucination and a factual error?
An AI hallucination occurs when a model generates a plausible-sounding but entirely invented statement or case citation. A factual error is a mistake based on real data that has been misinterpreted.
Hallucinations happen because models are probabilistic engines designed to predict the next word rather than retrieve verified truths. Verification is the primary defense against both, requiring that every citation be cross-referenced against sources before use.
Does using AI for client work violate attorney-client privilege?
Using AI does not automatically violate privilege if the firm uses secure and private infrastructure. The breach occurs when confidential information is uploaded to public AI platforms that store data for model training.
To maintain privilege, firms must use enterprise accounts with data-sharing settings disabled. Ensuring that no client information reaches the public domain is a mandatory step for any secure implementation.
Can I be sued if my law firm's AI makes a mistake?
You are responsible for all work product produced by your firm, whether created by an associate or an AI system. Clients expect professional work and courts demand accurate filings.
If an AI error results in a bad legal outcome, the firm faces the same malpractice risks as it would for a manual drafting error. Diligent oversight and review processes are the most effective ways to mitigate this professional liability.
How is the cost of a custom AI solution for a law firm determined?
Costs are scoped based on the complexity of the internal workflows and the volume of data involved in the integration. Professional firms evaluate the cost by looking at the specific manual labor hours saved through automation.
Engagements often start with an assessment of existing technical constraints and data architecture. A tailored system avoids the bloat of large, generic platforms by focusing on the specific tasks that provide the highest return on time invested.
Executive Summary
The widely repeated claim is wrong in an important detail. In Mata v. Avianca, the lawyers submitted a brief citing cases that ChatGPT had invented, and the court sanctioned them. They were not disbarred. The distinction matters, because the actual failure was not using AI: it was filing citations nobody verified, which breaches a duty that predates the technology entirely. The model produced plausible case names because plausibility is what it optimises for, and no step in the workflow checked them against a real reporter. Firms using AI safely have not banned it. They have made verification a named person's job before anything is filed.
What Should You Do Next?
Check whether your firm has a written rule naming who verifies citations before a filing goes out, and whether it covers AI-assisted drafts. Most do not yet. AiBuildrs has built more than 200 systems with review steps designed in. Discuss where the checks belong.
People Also Read
About the Author
Jerry Jariwalla is the founder of AiBuildrs and creator of the Growth Signal Intelligence framework. With over 22 years in digital marketing and multiple successful business exits, Jerry has spent the past decade leading AI implementation programs for mid-market businesses across professional services, recruitment, membership organizations, and traditional industries.
Expertise: AI Strategy, AI Implementation, Workflow Automation, Custom AI Development, Voice AI, Offshore Engineering, B2B Sales Intelligence, Mid-Market AI Adoption
Connect: LinkedIn
Disclaimer: This content is for informational purposes only and does not constitute professional business or technology advice. ROI outcomes vary based on industry, existing systems, and implementation commitment. Contact AiBuildrs for a consultation regarding your specific situation.
AI Buildrs
Written by the AiBuildrs team. We identify operational inefficiencies and build custom AI infrastructure to fix them permanently. Learn more about AiBuildrs →