Alexander v. Baas: Vexatious Litigant Declares AI Fabricated Quotations in Multiple Briefs

Background: Medi-Cal Benefits, Federal Preemption, and a Vexatious Litigant Label

Michael T-Alexander, a permanently disabled Medi-Cal beneficiary, filed suit in Santa Cruz County against the directors of California’s Department of Health Care Services and Department of Social Services, alleging his benefits were not properly transferred when he moved counties. He had previously filed a related lawsuit in 2022, and the 2023 petition was filed while that earlier action remained pending.

T-Alexander’s core legal theory was that California’s vexatious litigant statutes (CCP §§ 391–391.7) were unconstitutional because they violated the federal Supremacy Clause and were preempted by federal disability rights law. This argument was novel but legally thin — the court summarily rejected it. The trial court found T-Alexander to be a vexatious litigant under CCP § 391.1 and imposed a prefiling order under § 391.7.

T-Alexander appealed, and the Court of Appeal affirmed on all grounds. But the more notable aspect of the published opinion is not the vexatious litigant holding — it is the court’s documented discovery of multiple fabricated quotations in T-Alexander’s briefs.

The AI Issue: Fabricated Quotations Across Multiple Briefs

The court identified at least two instances in T-Alexander’s briefs where he quoted from real cases — but the quotations did not appear anywhere in those cases.

The most striking example: T-Alexander cited People v. Superior Court (Humberto S.) (2008) 43 Cal.4th 737, 750, for the following quotation: “a judge may not act on personal knowledge of facts not presented in evidence or conduct an independent investigation of facts not in evidence.” The court searched the case thoroughly and found that no such quotation appears in that case — or in any other case the court was able to locate. The language appears to be entirely fabricated.

The court noted this was not an isolated incident: T-Alexander’s opening brief contained “another such instance” that was also noted but not detailed in the opinion, and his reply brief “contain[ed] numerous similar fabricated quotations.” The record also showed that T-Alexander had acknowledged in the trial court that he had included inaccurate citations in prior briefing — suggesting awareness of the pattern, which mirrors the post-warning recidivism seen in cases like Burches v. Equifax.

The court issued a formal caution — noting that this conduct “violates the California Rules of Court and Code of Civil Procedure, notwithstanding the fact that he is unrepresented.” It cited CPC Rule 8.204(a)(1)(B) (requiring citations to legal authority for all assertions of law) and CCP § 128.7(b)(2) (requiring legal contentions to be warranted by existing law or a nonfrivolous argument).

Holdings

1. Vexatious litigant order affirmed. T-Alexander met the statutory criteria: he had filed multiple unmeritorious lawsuits, engaged in frivolous litigation conduct, and warranted a prefiling order to protect courts and opposing parties.
2. Federal preemption argument rejected. California’s vexatious litigant statutes do not violate the Supremacy Clause and are not preempted by federal disability rights law.
3. Fabricated quotations documented in published record. The court took the significant step of documenting specific fabricated quotations in a published opinion, creating a permanent appellate record of T-Alexander’s conduct.
4. Warning issued despite pro se status. The court explicitly stated that fabricating quotations violates the Rules of Court and CCP regardless of whether the litigant is represented — reinforcing the universal verification duty.

“T-Alexander’s reply brief contains numerous similar fabricated quotations. The record also shows that T-Alexander acknowledged having included inaccurate citations in briefing in the trial court. We caution appellant that such conduct violates the California Rules of Court and Code of Civil Procedure, notwithstanding the fact that he is unrepresented.”

— California Court of Appeal, Sixth District, April 14, 2026

India Angle: Fabricated Quotations from Indian Precedents

The Alexander v. Baas variant of AI hallucination — inventing a plausible-sounding quotation from a real case — is especially dangerous in Indian litigation, where judicial quotation of Supreme Court and High Court precedents is central to legal argument. Indian advocates frequently quote directly from SCC-reported judgments; if AI fabricates a quotation and attributes it to a real landmark case, Indian courts are less likely to spot it without a full-text search of the original judgment.

Relevant Indian Law

• Contempt of Courts Act 1971, Section 2(c): Wilful misrepresentation of facts before a court — including fabricating a quotation from a precedent — constitutes criminal contempt as it scandalises the administration of justice. If the court discovers that a quoted passage does not exist in the cited judgment, this is a direct misrepresentation to the court.
• BCI Rules of Professional Conduct, Rule 9 and Rule 22: Under Rule 9, an advocate shall not act in a manner tending to mislead the court. Under Rule 22, an advocate who has misled the court must promptly correct the record. Fabricating judicial quotations violates both rules and can lead to proceedings under Advocates Act 1961 Section 35 before the BCI.
• Indian Evidence Act 1872, Section 57: Courts may take judicial notice of the contents of Indian law reports. A fabricated quotation from a Supreme Court judgment — if read against the actual judgment — would constitute a material misstatement of the law to the court, inviting contempt proceedings.

Three Practical Tips for Indian Practitioners

1. Never quote from a judgment without reading the quoted passage in the original. AI tools generate legally plausible-sounding quotations that may have never appeared in the cited case. Before including any direct quotation from a judgment, locate the passage in the original SCC Online, Manupatra, or Indian Kanoon text and visually confirm it exists. Copy from the original — never from the AI’s output.
2. Be especially careful with Supreme Court five-judge bench and Constitutional Bench quotations. AI models frequently attribute aphoristic legal statements to landmark judgments like Kesavananda Bharati, Minerva Mills, or Maneka Gandhi that do not appear in those cases. These are high-visibility citations; courts and opposing counsel know these judgments well and will spot fabrications.
3. If you discover a fabricated quotation in your brief before filing, fix it. The moment you find an AI-generated quotation that cannot be verified, remove it and substitute a genuine quotation from a verifiable source, or omit the quotation entirely. Never rationalise that “the court probably won’t check.” Indian High Courts and the Supreme Court regularly read cited authorities carefully.

Quick Takeaways

• AI tools do not just fabricate case names — they fabricate specific quotations attributed to real cases, producing text that is legally plausible but entirely invented.
• Courts are now actively searching cited cases for quoted passages — and documenting in published opinions when those passages cannot be found.
• A prior acknowledgment of citation errors (as T-Alexander made in the trial court) and then continuing to fabricate in the appeal is a pattern courts find particularly troubling.
• Vexatious litigant status and AI hallucination may co-occur: a litigant who repeatedly files may also repeatedly use AI for filings, compounding both patterns.
• In India: fabricating a quotation from an Indian Supreme Court case is serious misconduct under both the Contempt Act and the BCI Rules — no defence of pro se status or AI error will insulate an advocate from consequences.

Deep Dive: Fabricated Quotations — The Most Dangerous AI Hallucination Variant

Of all the ways AI hallucinations manifest in legal filings, fabricated quotations are the most dangerous — and the hardest to catch. A fabricated case name is obviously suspect if the case does not appear in a citator search. A fabricated statutory citation fails at the first check in the official code. But a fabricated quotation attributed to a real, existing case requires the reader to open the case, locate the passage, and confirm it exists — a step that neither busy judges nor opposing counsel reliably perform for every citation in every brief.

The Alexander v. Baas pattern has appeared in earlier AI cases too. In Fecteau v. Safety-National, fabricated quotes triggered a filing injunction. In Mata v. Avianca (the seminal 2023 case), Judge Castel specifically highlighted that the invented cases came with invented quotations designed to make them sound real. The fabricated-quotation variant is a signal that the AI was prompted not just to find authority but to quote it — amplifying the hallucination from citation level to textual level.

For Indian practitioners, the risk is particularly acute because Indian legal writing is heavily quotation-dependent. Judgments commonly quote from earlier Supreme Court decisions at length. AI models trained on these conventions will naturally generate lengthy quotations when asked to draft Indian legal arguments. Each such quotation must be independently verified before it enters a brief or pleading.

The permanent appellate record aspect of this case is also significant. Because the California Court of Appeal published its opinion with the specific fabricated quotation documented, this is now a matter of public record. T-Alexander’s conduct will appear in any future background check, vexatious litigant search, or legal research into his cases. Indian courts increasingly publish judgments on Indian Kanoon and SCC Online; similar conduct by an Indian advocate would be similarly immortalised in the official law reports.