Court Reprimands Lawyer for AI Hallucinations in Briefs
The Eleventh Circuit affirmed dismissal of a workplace-policy case and reprimanded attorney Anthony Sabatini for appellate briefs containing fabricated case citations produced with AI. The court's published opinion says the filings included nonexistent authorities and that an attempted correction introduced additional false citations. The decision reinforces a simple professional rule: lawyers remain responsible for every authority submitted under their names, regardless of which tool drafted or researched the text. For legal and AI teams, the failure was not only model hallucination; it was the absence of source verification before filing. Citation-grounding systems, human review, and direct checks against official case databases are necessary controls when generative tools enter legal research or drafting.
This opinion turns a familiar AI warning into an institutional record: the professional using the tool, not the model vendor, carries the duty to verify a court filing. Generative assistance may speed research or drafting, but it does not change the evidentiary status of a citation. A plausible case name is still unusable until a reviewer locates the authority, confirms the proposition, and checks that it remains good law.
What happened
The Eleventh Circuit affirmed dismissal in Akerlund v. Atlas Air and criticized plaintiffs' counsel Anthony Sabatini for briefs containing fabricated authorities. The court's published opinion describes multiple false or hallucinated citations and says the attorney's attempted withdrawal did not match all of the problematic authorities; substitute citations were also fabricated. The Volokh Conspiracy report highlighted the court's language and the unusual sequence of errors. The legal outcome and reprimand are grounded in the official opinion, which is the authoritative record. Commentary adds context but does not replace the decision.
Technical context
Citation hallucination can arise when a language model generates text that resembles legal writing without retrieving from a controlled corpus. Even retrieval-augmented systems can fail if they return the wrong jurisdiction, an outdated decision, or a source that does not support the proposition. Verification therefore needs multiple checks: the citation exists, the quoted or paraphrased rule appears in the authority, the procedural posture is understood, and later history has not undermined the case. A model's confidence or polished prose is not evidence.
For practitioners
Legal teams should require a source packet for every AI-assisted filing, with links or copies from official or trusted databases. Reviewers should verify citations independently rather than asking the same model to check its own output. Firms can use automated citation parsers as a screen, but final responsibility should remain assigned to named lawyers. Matter policies should also define which client information may enter external tools, how prompts are retained, and when AI assistance must be disclosed under court or professional rules.
What to watch
Courts are developing different responses to fabricated citations, including reprimands, sanctions, education requirements, and filing restrictions. Watch whether this opinion influences local rules or disciplinary proceedings and whether legal-software vendors add stronger source-binding features. The broader adoption signal will be whether organizations measure verified research quality rather than drafting speed. AI can still assist legal work, but only workflows that make source inspection unavoidable can reduce this recurring failure mode.
Key Points
- 1An appellate panel affirmed dismissal of the suit for lack of personal jurisdiction and failure to state a claim.
- 2The opinion found counsel presented briefs with fabricated case citations after relying on an artificial intelligence tool for research and drafting.
- 3The court concluded the attorney violated ethical obligations to clients and the court by outsourcing legal work to an AI algorithm.
Scoring Rationale
A federal appellate opinion creates a clear, public precedent-like warning about professional responsibility for AI-generated legal errors. The event is important for legal AI governance, though its direct scope is one case and one attorney's filings.
Sources
Primary source and supporting public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
