Galang v. Canada (Attorney General): BC Supreme Court Orders Verified AI Disclosure After Hallucinated Citations | Advocate Prakhar

Case at a Glance

Full Citation	Galang v. Canada (Attorney General), 2026 BCSC 564 (CanLII), S258586 (BCSC Mar. 31, 2026)
Court	Supreme Court of British Columbia
Date	March 31, 2026
Category	AI Hallucination / Verification Order
Jurisdiction	Canada (British Columbia)
AI Tool Used	Inferred generative AI (hallucinated cases in pleadings noted by court)
Outcome/Sanction	Legal basis section struck and limited to 10 pages; costs of $250 awarded; mandatory AI verification order imposed

Background

Ronuel “Alex” Galang, appearing as a self-represented litigant, filed a Notice of Civil Claim (NOCC) against the Attorney General of Canada and Fraser Health Authority spanning 90 pages. The NOCC’s legal basis section (Part 3) was approximately 62 pages long and was brought to the court’s attention by defence counsel as containing perceived use of generative AI, including references to non-existent cases that appeared to be AI hallucinations. Justice Dion heard an application to strike Part 3 in October 2025 and entered an order striking it with leave to amend within 15 pages, paying $250 in costs. The March 2026 reasons by Justice Veenstra address subsequent proceedings after Galang filed his Amended Notice of Civil Claim (ANOCC).

The AI Issue

The court had to determine what procedural conditions are appropriate when a self-represented litigant is permitted to continue litigation after submitting AI-hallucinated pleadings. Rather than barring further filings outright, the courts imposed a combination of length restrictions and a positive verification obligation: if the plaintiff uses AI to assist in drafting, he must verify all cited cases and court rules before filing. This places the verification duty explicitly on the litigant rather than leaving it as an implied professional obligation.

What the Court Decided

• Part 3 (the 62-page legal basis section) of the original NOCC was struck with leave to amend, limited to 10 pages [structural remedy for AI-inflated pleadings].
• The plaintiff was ordered to pay $250 in costs to the Attorney General of Canada payable forthwith [monetary sanction].
• A specific AI verification order was imposed: “If the plaintiff uses artificial intelligence to assist drafting her material in this action, she is required to ensure the Court cases or Court Rules she refers to exist prior to relying on such Court cases or Rules prior to filing” [mandatory verification condition].
• References to non-existent cases in the NOCC were identified as constituting AI hallucinations that impose unacceptable costs on defendants and the court [harm to judicial process confirmed].
• The Amended NOCC was reduced from 90 to 32 pages in compliance with the order, with Part 3 reduced to 6.5 pages [substantial compliance achieved].

“If the plaintiff uses artificial intelligence to assist drafting her material in this action, she is required to ensure the Court cases or Court Rules she refers to exist prior to relying on such Court cases or Rules prior to filing such material with the Court.”

— Justice Dion, Supreme Court of British Columbia, Order dated October 3, 2025 (summarised in Galang v. Canada, 2026 BCSC 564)

The India Angle

Indian Law Equivalent

India has no specific court rule equivalent to the BC verification order yet, but the mechanism for imposing such conditions exists. Under the Supreme Court Rules, 2013 (Rule 2 of Order XXXVII) and High Court rules on pleadings, courts may impose conditions on the filing of amended pleadings, including requiring that they be limited in length and verified. Under Section 151 CPC (inherent powers) and Order VI Rule 16 (striking out pleadings), courts can strike out portions of pleadings that are frivolous, scandalous, or likely to embarrass the fair trial — a power fully applicable to AI-hallucinated legal basis sections.

Bar Council Rules

BCI Rule 14 (no false statements), Rule 15 (no disrepute), and Rule 22 (dignity of court) apply to advocates who assist self-represented litigants or who draft pleadings containing AI-generated content without verification. The Galang-style verification order, if adopted by Indian High Courts, would most naturally sit as a condition imposed under Order VI Rule 16 CPC or under the court’s directions for case management in complex civil litigation. The Law Commission of India should consider recommending such verification conditions as part of a broader AI governance framework for civil procedure.

Practical Advice for Indian Advocates

• When your client’s pleadings are struck or returned for amendment due to AI citation errors, treat the amendment opportunity as a full quality review — use authenticated databases for every citation in the amended document, not just those previously flagged.
• Length restrictions in AI-generated pleadings are now an emerging judicial tool — courts are using page limits as a practical constraint on AI-inflated submissions; draft concisely and substantively from the outset.
• Verify all statutory and rule references, not only case citations — AI tools also hallucinate rule numbers and statutory provisions, as the Galang court noted with respect to “Court Rules.”

Quick Takeaways

• Canadian courts impose mandatory AI verification orders as a continuing litigation condition.
• AI hallucinations inflate pleading length — courts respond with page limits and restructuring orders.
• Statutory and rule references must be verified, not just case citations, in AI-assisted filings.

---

Deep Dive: The Verification Order as a New Judicial Tool for AI Governance

The most innovative aspect of the Galang proceedings is not the monetary sanction ($250) or the requirement to refile with length restrictions — both of which are familiar court tools — but the positive verification order imposed by Justice Dion as a condition of continued litigation. This order explicitly names generative AI as the anticipated tool and requires the plaintiff to verify all cited cases and court rules exist before relying on them in any future filing. It is, in effect, a court-mandated quality assurance protocol applied at the individual litigant level.

This approach represents a middle path between the punitive focus of monetary sanctions (Fogarty’s $2,500, Mata v. Avianca’s $5,000 per attorney) and the structural focus of court-wide disclosure requirements (Essandoh’s certification framework). The verification order is targeted and proportionate: it does not prohibit AI use, does not require disclosure of which AI tool was used, and does not extend to other litigants in other cases. It simply requires the specific litigant who demonstrated AI hallucination problems to verify accuracy before filing — which is, of course, what every litigant should do regardless of court order.

The broader context of the Galang case — a pro se plaintiff suing two powerful defendants (the federal Attorney General and a provincial health authority) in relation to events during a hospitalisation — makes the AI verification order particularly significant. Self-represented litigants in complex civil cases are precisely the population most likely to over-rely on AI tools without adequate verification, because they lack both the legal training to evaluate AI output and the institutional resources (legal databases, supervising colleagues) to cross-check it. Courts that impose verification orders as a litigation condition are effectively educating these litigants about the nature of their responsibility, rather than simply punishing failures after the fact.

For India, the Galang precedent suggests a useful role for National and State Legal Services Authorities (NALSA and SLSAs) and District Legal Aid Committees. As these bodies increasingly use AI tools to assist self-represented litigants in preparing pleadings and applications, they need verification protocols that match the standard being established by courts in Canada, the US, and the UK. A verification order equivalent to Justice Dion’s, built into the NALSA AI assistance workflow, would both protect self-represented litigants from having their filings struck and protect courts from the burden of AI-contaminated proceedings.