Florida Appeals Court: AI Legal Garbage Clogs Case Files
The Florida Fourth District Court of Appeal warned in a divorce case ruling that materials submitted by lawyers and pro se litigants generated by artificial intelligence are clogging case files and threatening the operation of the judiciary. Judge Robert Gross stated that the misuse of AI submissions jeopardizes orderly justice; lengthy, unfocused, and nonsensical documents were directly referred to as "AI slop." The panel ordered family lawyer Jaclyn Soroka to explain within ten days why she should not be sanctioned and considered referring her to the state bar association.
The focus of the case has shifted from "fabricating precedents" to "lack of independent judgment." The panel noted that the cited cases are real, the language is confident, but the arguments resemble tool assembly rather than a trained lawyer's reasoning. Gross wrote that even a supercomputer would struggle to extract meaning from some of the writings; the documents resemble those produced by habitual pro se litigants, seemingly intended to delay proceedings. The court suspects that some untenable arguments stem from rough trial transcripts being miswritten and required her not to use AI in her responses.
The same court system has previously dealt with hallucinated citations multiple times. In the Gouveia case, Appeals Judge Melanie May described the technology as a miracle of the era, warning that without approval it could become a judicial plague; concurring Judge Johnathan Lott stated that lawyers are repeat offenders, and sanctions are necessary to make them remember, while pro se litigants continue to submit a stream of chatbot-generated nonsense. Broward Chief Administrative Judge Carol-Lisa Phillips noted that documents that used to be five to ten pages long now often reach ninety pages, requiring judges to read page by page, with searching for a nonexistent case consuming their schedules. At a bar association meeting, a federal magistrate stated that pro se submissions statewide have increased by about 10% to 15%, with each submission growing from a few handwritten pages to ninety pages.
On May 28, the Florida Supreme Court enacted a self-amendment that took effect on June 15: signatories must declare that the legal basis cited exists and is accurately referenced, with violators facing admonishment, contempt, document deletion, dismissal of proceedings, costs, and attorney fees as sanctions. The court replaced individual circuit courts' AI disclosure orders with a unified state regulation. The 11th and 17th circuits had previously required lawyers to confirm citations, which the state high court subsequently superseded. Tracking databases show that in recent years, over 1,500 related court orders have been issued globally, with at least 54 from Florida courts.
In market mechanisms, this represents a quota war over judicial hours against generated texts. Buyers are lawyers and pro se litigants looking to fill motions with cheap drafts; sellers are model companies and legal research subscriptions. Funding has shifted from hourly writing to token-based submissions reviewed for free by judges. Beneficiaries include retrieval companies that can verify citations live, compliance training, and disciplined lawyers willing to take on sanctionable work; those under pressure are court clerks' schedules and drafting shops that charge by page. Sanctions shift external costs from judges' desks back to signatories. What is truly being priced is not the intelligence of the model, but who must work overtime for ninety pages of garbage.
ABAB AI Insight
Florida has taken the lead with administrative orders from various circuit courts, later consolidated by the state Supreme Court into a signature guarantee. The institutional choice is clear: there is no mandatory disclosure of "whether AI was used," only a guarantee that "the case is real." The Fourth District Court of Appeal has pushed the line further: it is not enough for the case to be real; the reasoning must resemble that of a human. This establishes a second charge after hallucinations—absence of professional judgment. Family courts are particularly hard hit due to fragmented procedures, dense motions, and numerous pro se litigants, making them ideal for models to turn conflicts into endless pages.
The capital path shifts the downstream costs of lawyers' hourly fees onto the public judicial budget. Models drive the marginal cost of drafting close to zero, while the marginal cost of court reading remains one person per page. Ninety pages versus five pages represents an asymmetry between computational power and drafting. The sanctions menu provided by the state Supreme Court (deleting documents, dismissals, fines) gives judges a tool for recovering costs. Legal tech companies sell drafting while also selling verification; the same round of technology both creates blockages and offers solutions. Large plaintiffs have invested hundreds of millions in proprietary systems, while directly generating outputs from chatbots follows two completely different chains of responsibility.
This is analogous to the era of fax litigation letters, bulk-format motions after electronic filing, and discovery processes being abused to wear down opponents. This time, generators allow "lengthy but unfocused" to be infinitely replicated. The industry stage is regulatory change: professional responsibility expands from "what you wrote" to "whether you exercised judgment." At the federal level, judges have already proposed similar verification rules. Whoever first establishes verification as a mandatory infrastructure will take the next bill in legal software subscriptions.
Structural judgment is the overlay of technological substitution and regulatory change. The mechanism is an explosion of text supply, while judicial attention has not expanded. Pricing power shifts from those who can write the longest documents to those who can prove that every citation and every segment of reasoning has been reviewed by a human. Courts are re-pricing attention through sanctions.