Eagle Township v. VanAlstine: AI Hallucinations in Court’s Own Opinion — Westlaw Flags Michigan Appellate Decision

Background: A Pole Barn, a Zoning Permit, and an IZO Violation

Shaun VanAlstine owns a 2.43-acre property in Eagle Township, Michigan, located in an agricultural zoning district. The property predates the township’s Interim Zoning Ordinance (IZO) and is a legal nonconformity because it falls below the 10-acre minimum lot size for the agricultural district. In early 2024, VanAlstine planned to build an addition to an existing pole barn on the property for storing work vehicles and equipment.

VanAlstine’s wife contacted the county building official informally about setback requirements. Based on that informal guidance — which came without knowledge of the property’s specific dimensions or zoning status — VanAlstine proceeded with construction without obtaining the required township zoning permit. The township discovered the construction, found it violated multiple IZO requirements (setback distances, maximum accessory structure floor area, and nonconforming structure enlargement restrictions), and brought suit for declaratory and injunctive relief.

The trial court granted summary disposition for the township, ordered VanAlstine to remove the structure, and awarded costs and attorney fees. VanAlstine appealed, arguing that he had relied in good faith on the building official’s guidance. The Court of Appeals affirmed on all grounds — the building official’s informal guidance did not estop the township from enforcing its zoning ordinance, and VanAlstine had not obtained the required approvals before constructing.

The AI Issue: Hallucinations in the Court’s Own Opinion

The substantive legal issues in Eagle Township v. VanAlstine are unremarkable. What places this case in the AI hallucination law corpus is the Westlaw Editor’s Note appended to the published decision:

“This decision contains discussion of citation references that are incorrect or do not actually exist. These invalid citations appeared in the original court opinion and have been preserved as written since they are part of the official record. Any links to these invalid citations have been removed.”

This is the critical distinction: the AI hallucinations here are not in a party’s brief. They are in the court’s own opinion. Westlaw identified citation references in the appellate court’s reasoning that are incorrect or nonexistent — likely generated by or incorporated from an AI-assisted drafting tool used by the court itself (or its law clerks).

This variant of AI hallucination is categorically more serious than party-brief hallucinations. When a party files a brief with a fabricated citation, the court and opposing counsel have a chance to spot and correct it. When the court’s own published opinion contains fabricated citations, those citations enter the official law reports, become citable precedent, and can be relied upon by future litigants — including, ironically, by AI tools that scrape legal databases for training data, creating a feedback loop of AI-generated error.

Holdings on the Zoning Merits

1. IZO violations established. The accessory structure violated Eagle Township’s setback requirements, maximum floor area limits (Condition Set Two, § 6.02.C.1), and the prohibition on enlarging nonconforming structures (§ 7.04.A).
2. Good faith reliance on informal guidance does not estop the township. The building official’s email response to an informal inquiry, given without knowledge of the property’s dimensions or zoning status, did not constitute an authoritative representation sufficient to estop the township from enforcing its IZO.
3. Costs and attorney fees affirmed. The trial court’s award of costs and attorney fees to the township was upheld.
4. Judicial AI hallucinations documented by Westlaw. The Westlaw Editor’s Note flags this decision as containing invalid citations that appeared in the court’s own opinion — a significant escalation from party-brief AI errors into the judicial drafting process itself.

“Editor’s Note: This decision contains discussion of citation references that are incorrect or do not actually exist. These invalid citations appeared in the original court opinion and have been preserved as written since they are part of the official record.”

— Westlaw Editor’s Note, Eagle Township v. VanAlstine, 2026 WL 1194667

India Angle: When the Court’s Own Judgment Contains an AI Error

The most alarming dimension of Eagle Township v. VanAlstine is the question it raises: what happens when Indian courts begin using AI drafting tools and their judgments contain fabricated citations? This is not a hypothetical. Indian courts at both the High Court and Supreme Court level have announced AI initiatives; judgment-drafting assistance tools are being explored.

Relevant Indian Law

• Review jurisdiction under CPC Order XLVII and Article 137: If an Indian court judgment contains an incorrect citation that materially affected the ratio, the aggrieved party may apply for review (High Courts) or curative petition (Supreme Court) on the ground of apparent error on the face of the record. A citation to a nonexistent case — if that citation was the basis of a holding — could constitute such an error.
• Advocates Act 1961 Section 35 and BCI Rules: While judicial AI errors are not directly governed by the BCI Rules (which apply to advocates, not judges), the practical consequence for advocates is significant: if you cite a judgment that contains AI-hallucinated reasoning, and your opponent discovers the underlying citation is bad, your argument built on that flawed authority collapses — with potential cost consequences under CPC Order XX.
• The Supreme Court AI Framework (2024): The Supreme Court’s e-Courts mission and the SUVAS/SUPACE projects have incorporated AI assistance in court management. Any AI tool used in drafting requires the same verification standard applied to party submissions. India has no formal equivalent of Westlaw’s Editor’s Note system — errors in Indian judgments, once published on Indian Kanoon or SCC Online, may not be flagged at all.

Three Practical Tips for Indian Practitioners

1. Always verify citations in court orders and judgments you intend to rely upon. Before citing a prior court order or judgment in support of your argument, independently verify that the citations within that judgment are accurate. If a judgment you are relying upon itself contains a fabricated citation, your argument may be built on an unstable foundation.
2. Raise objections to incorrect citations in court orders at the earliest stage. If you discover that a court order in your matter contains an incorrect or nonexistent citation, bring it to the court’s attention in your next filing or at the next hearing through a formal mention. Courts have inherent power to correct typographical and clerical errors in orders under Section 152 CPC; citation errors may fall within this power.
3. Maintain independent research files for all precedents you rely upon. Do not rely solely on what a court has summarised about a case — read the original. AI-assisted judicial drafting may produce summaries of cases that are accurate in their results but imprecise or inaccurate about specific passages or holdings. Your independent verification protects your client if the summarised authority is later found to be mischaracterised.

Quick Takeaways

• AI hallucinations are no longer confined to party briefs — they are appearing in court-generated opinions, as evidenced by Westlaw’s Editor’s Note on this Michigan appellate decision.
• When incorrect citations appear in a court’s own opinion, they enter the official record, cannot easily be corrected, and may be relied upon in future cases — creating a compounding error risk.
• Westlaw’s Editor’s Note system is currently the only systematic mechanism for flagging AI-impacted judicial opinions; no equivalent exists in India or most Commonwealth jurisdictions.
• The correct response to discovering a hallucinated citation in a court opinion you are relying upon is to independently verify the actual citation — not to cite the opinion’s version of the authority.
• In India: judgments are authoritative as written; an AI error in a judgment becomes part of the official record unless corrected through review, clarification, or curative petition proceedings.

Deep Dive: When the AI Hallucination Problem Moves Upstream to the Bench

Eagle Township v. VanAlstine is part of a small but growing category of AI hallucination cases where the error originates not in advocacy but in adjudication. Earlier cases in this category include Amparo Trejo v. Miguel Angel Amaya Hernandez (D. Maryland 2026), where Westlaw flagged AI hallucinations in the court’s own published opinion in a Special Immigrant Juvenile Status case. Both cases share the same disturbing characteristic: the AI error is baked into the official record at the judicial level, not the advocacy level.

The implications differ significantly depending on which side of the bench the error originates. A party’s fabricated citation can be challenged by the opposing party, struck by the court, and sanctioned. A court’s own fabricated citation cannot be challenged through normal adversarial means — it would require a motion for clarification, a reconsideration request, or a formal judicial error correction process. In federal courts, Local Rules vary widely on whether and how courts correct citation errors in published opinions.

For Indian jurisprudence, this problem has a particular dimension: India’s common law tradition treats reported judgments as authoritative on the principles they state, not merely on the citations they contain. An incorrect citation within an Indian court judgment is less likely to be treated as error-in-the-opinion and more likely to be treated as background noise — unless the incorrect citation was the basis of the ratio. Indian practitioners should be alert to this possibility as AI-assisted judicial drafting becomes more widespread.

The long-term systemic risk is a feedback loop: AI tools are trained on legal databases; if AI-hallucinated citations enter court opinions and those opinions are then indexed in Westlaw, Lexis, SCC Online, or Indian Kanoon, future AI tools will learn from those hallucinated citations and reproduce them in future outputs. Westlaw’s Editor’s Note system is a partial safeguard — but it requires human detection of the error in the first place. The deepest lesson of Eagle Township is that AI verification standards must apply to all legal writing — advocacy, judicial opinions, and everything in between.