AI fabrications keep hitting courts as legal tools spread
Regulators and courts have flagged a rising number of legal filings containing AI-generated fabrications, including invented citations and case law. The Ontario Law Society fined a lawyer $31,150 for using Grok to draft a factum full of fake legal principles, while Sullivan & Cromwell had to alert a New York bankruptcy court after a filing contained 42 fabricated AI citations. Canadian courts logged a sharp rise in such cases, from seven in 2024 to 86 in 2025 and 39 more in early 2026.
GoKawiil's interpretation of the reporting above, not reported fact.
The pattern suggests that even well-resourced, elite law firms are not immune to AI-generated errors slipping into official filings, which could undermine trust in legal AI tools industry-wide. Tribunal adjudicators' comments—that AI 'does not appreciate nuance' and prefers giving an answer over admitting ignorance—point to a structural limitation rather than a one-off failure, which may slow adoption of AI in high-stakes legal work despite vendors' ambitions.
- A Canadian lawyer was ordered to pay $31,150 after AI-generated fabrications appeared in a legal factum.
- Sullivan & Cromwell, a top US law firm, had to disclose 42 AI-fabricated citations in a bankruptcy filing.
- Reported fabrication cases in Canadian courts jumped from 7 in 2024 to 86 in 2025, with 39 more by early 2026.
Source: engadget.com — Luke James, 2026-09-29
Published there as: “The AI industry wants models to assist in legal battles, but will they help?”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.