A federal judge has dismissed with prejudice a whistleblower lawsuit accusing Los Angeles Film School of inflating job-placement results to protect accreditation and access to federal student loans. The ruling also imposed a $5,000 sanction on attorney Mark Kleiman after a 76-page opposition brief filed in the case relied on generative AI and included a citation to a case that does not exist.

The decision is consequential for two separate reasons. First, it ends this particular False Claims Act case, meaning the same claimants cannot simply refile the dismissed claims. Second, it is another clear demonstration that AI output cannot substitute for a lawyer’s verification of legal research, factual claims, and court filings. A lengthy document, polished language, and legal-sounding case citations are not evidence that a brief is accurate.

What the lawsuit alleged

The suit was brought in 2024 by Dave Phillips, a former vice president of job placement at Los Angeles Film School, and Ben Chaib, a former vice president of admissions. Their complaint alleged that the school paid outside vendors to employ students in very short-term jobs, including alleged two-day positions, so the school could report stronger job-placement statistics.

Among the entities named in the allegations was Ivar Music Group. Phillips and Chaib contended that artificially enhanced placement figures were used to help preserve accreditation and eligibility for federal student-loan funding. The complaint also alleged that many graduates did not go on to work in film or music, citing an internal document said to state that most graduates reported yearly income of $0 to $5,000 in their field of study.

Those were allegations in the complaint, not findings that were established in this case. The court did not adjudicate the alleged fraud on its merits after concluding that much of the alleged conduct was outside the applicable six-year statute of limitations.

Los Angeles Film School had previously paid $1 million in 2020 to resolve a Department of Education audit involving this and other issues. The former executives alleged that the school had not fully disclosed the extent of the alleged conduct to department investigators. Their suit also named CEO James “Bill” Heavener and included allegations involving Full Sail University in Winter Park, Florida, which Heavener owns as well. The judge found that the claimants had not supplied specific details supporting the allegations concerning Full Sail.

Why the dismissal matters

A dismissal with prejudice is a final procedural outcome for the claims addressed by the order. In practical terms, it bars the plaintiffs from bringing the same case again. That differs from a dismissal without prejudice, where a court may leave room for a revised complaint or a later refiling.

The statute of limitations issue was central. A statute of limitations sets the time window in which a plaintiff must file a claim. Here, Judge Stanley Blumenfeld concluded that a substantial portion of the alleged conduct arrived too late under the six-year limit that applied to the claims. Deadlines can decide a case even when the underlying accusations are serious, because courts require claims to be brought within legally defined periods.

The case was filed under the False Claims Act, a federal law that allows private parties, often called whistleblowers or relators, to sue on the government’s behalf over alleged fraud involving government money. If such an action succeeds, whistleblowers can potentially receive a share of a recovery. The mechanism is particularly significant in sectors that rely on federal funding, including schools whose students use federal loans.

But the existence of a False Claims Act theory does not remove ordinary litigation requirements. A complaint must provide sufficient factual detail, the claims must be timely, and the parties’ filings must accurately identify the law they say supports their arguments.

The AI brief and the court’s sanction

The legal filing became an issue of its own. In April, Kleiman submitted a 76-page brief opposing dismissal. It cited Darwich v. Department of Economic Security, a purported case that did not exist. The brief also cited two real decisions that did not support the arguments attributed to them. One of those decisions had been authored by Judge Blumenfeld himself.

Phillips later acknowledged that he had drafted the document with ChatGPT and provided it to Kleiman for review. Kleiman filed the document with a disclaimer intended to distance himself from its preparation beyond filing it to meet a deadline. He later told the court that his relationships with Phillips and Chaib had deteriorated. One week after the filing, he told opposing counsel that the clients had prepared it with AI and that he intended to withdraw.

The judge nevertheless treated the attorney’s responsibility as clear. The sanction order faulted Kleiman for a sustained pattern of neglect and violations of court orders and ethical obligations. The court emphasized that Kleiman knew nonlawyer clients had used AI, yet filed a deficient brief that he had not fully read and whose citations he had not verified.

“Kleiman filed an obviously deficient brief that he had not even fully read and for which he did not check whether the citations were to real cases — let alone that they supported the brief’s assertions,” the judge wrote.

The ruling also described the AI-generated opposition as containing almost no meaningful citations to legal authority or the complaint, while making arguments that appeared disconnected from the False Claims Act and relevant Ninth Circuit law.

Generative AI systems predict and assemble language. They are capable of producing text that appears coherent and authoritative even when its factual foundation is missing or wrong. In legal work, that can surface as a hallucinated citation: a case name, court decision, quotation, statute, or proposition that is fabricated, misdescribed, or unrelated to the point for which it is offered.

The danger is not limited to totally invented cases. The filing here illustrates a second problem: a citation may point to a real case but still fail to support the argument attached to it. Verification therefore involves more than confirming that a case appears in a database. A responsible review must check the full decision, the court that issued it, whether it remains valid authority, and whether its actual holding applies to the current dispute.

That level of review is a lawyer’s work, not a formatting exercise. Courts depend on adversarial briefing: each side presents authority accurately so the judge can evaluate the law and the record. A made-up authority wastes time, compromises the filing party’s credibility, and can force the court and opposing side to untangle claims that should never have been submitted.

AI can be useful for preliminary organization, plain-language explanations, or drafting ideas, but this ruling underscores the hard boundary in litigation: the filing attorney remains accountable for every representation made to the court. The practical standard is not whether an AI tool produced the passage; it is whether a lawyer independently confirmed that the passage is accurate, legally relevant, and properly supported.

The broader technology conversation is also reaching large companies outside courtrooms, including in coverage of AI infrastructure investment and workforce changes. This case is a narrower but important reminder that expanding use of the technology does not dilute professional duties in regulated or high-stakes settings.

Why the claimants were not fined

Judge Blumenfeld said Phillips and Chaib were not blameless in the AI episode, particularly because they did not obtain replacement counsel sooner as their relationship with Kleiman deteriorated. However, the court chose not to sanction them because they are not lawyers.

That distinction is important. Nonlawyer clients can contribute facts, review drafts, and communicate their positions to counsel. They should not assume that using a chatbot to generate a legal brief produces court-ready work. Yet the court’s order put the professional accountability for filing and validating the brief on the attorney who submitted it.

For anyone involved in a legal dispute, the practical lesson is straightforward: do not treat AI-generated text as verified research. Preserve underlying documents, identify the actual source for each factual assertion, and ensure that counsel checks each citation and legal argument before anything is filed. A deadline does not make an unreviewed document safe to submit.

A case about both timing and process

The dismissal does not turn the former executives’ allegations into proven falsehoods, nor does the prior Department of Education audit resolution itself establish the broader allegations presented in this lawsuit. Instead, the order resolved the case before the court based on its procedural and legal deficiencies: the lateness of much of the alleged conduct, insufficiently detailed claims relating to Full Sail, and an opposition filing that the judge found fundamentally unreliable.

For education businesses that market career outcomes, job-placement reporting remains a high-stakes subject because students, accreditors, and federal aid programs may rely on it. For lawyers and claimants, the ruling provides an equally direct warning about litigation conduct. A potentially consequential complaint can be derailed by untimely claims and by a failure to subject AI-assisted material to ordinary professional scrutiny.