Mediation & Arbitration / 2 min read
What happened?
A California appellate court recently imposed a $10,000 sanction on a Los Angeles-area attorney after finding that his opening brief contained fabricated quotations attributed to case law. The court concluded that 21 of 23 quoted passages were not real and emphasized a simple rule: no filing should include citations the submitting lawyer has not personally read and verified. The decision is among the largest AI-related sanctions by a California court and arrives amid a broader judicial and regulatory push to curb unreliable uses of generative AI in legal proceedings.
Why it matters?
This ruling reflects a growing, nationwide judicial skepticism toward unsupervised use of generative AI in litigation. Courts are not banning AI outright. Instead, they are reaffirming lawyers’ gatekeeping duty to ensure accuracy, candor, and compliance with procedural rules. Judges have sanctioned lawyers for “bogus AI-generated research,” cautioned against treating large language models as authoritative, and warned that human verification is non-negotiable. Beyond litigation risk, these incidents can damage credibility with courts and clients and trigger ethical or disciplinary consequences.
The regulatory backdrop
In California, the Judicial Council has directed courts to either prohibit generative AI or adopt formal use policies, with compliance deadlines that underscore the urgency. The State Bar is evaluating whether to update professional conduct rules to address emerging AI issues. Similar efforts are underway around the country and abroad, reflecting a fastchanging governance landscape that will shape how lawyers and legal teams deploy AI.
What this means for businesses and in‑house teams
For corporate legal departments, the message is clear: AI can speed drafting, summarization, and document review, but it cannot replace professional judgment, source validation, or confidentiality safeguards. Vendors are rapidly integrating AI into research, drafting, and eDiscovery tools. Even so, courts expect a human lawyer to stand behind every citation, quotation, and representation of fact or law. Where outside counsel or alternative legal service providers leverage AI, clients should expect written assurances regarding verification standards, privilege and confidentiality protections, audit trails, and model governance.
Practical recommendations
Organizations that have adopted or are considering AI-enabled legal workflows should formalize guardrails now.
- Establish an AI use policy covering permissible use cases, required human review, and prohibited inputs, with special attention to client confidences, trade secrets, and personally identifiable information. Specify that no AI output may be filed or sent externally without human verification against primary sources.
- Require documented cite-checking, quotation verification, and factual corroboration for any AI-assisted legal analysis. Treat LLM outputs as drafts or brainstorming aids, not authorities.
- Vet vendors for zero data retention options, encryption, access controls, clear trainingdata boundaries, and auditability. Align contract terms with your confidentiality, data residency, and incident response standards.
- Train legal teams and business stakeholders on AI capabilities and limitations, including hallucination risk, bias, and provenance. Pair training with spot checks and periodic audits.
- Align billing practices and engagement letters with AI-enabled workflows, including disclosure expectations, supervision standards, and quality controls for outside counsel.
Bottom line
Generative AI is becoming a durable feature of legal work, but courts are drawing a hard line on accuracy and accountability. The recent California sanction is a warning, not a ban: use AI to accelerate work, not to substitute for the lawyer’s duty to verify. Clients should insist on clear policies, technical safeguards, and demonstrable human oversight across all AI-influenced legal outputs.