The Firm
Harrison & Cole LLP is a 28-attorney regional law firm with $12 million in annual revenue. With practice areas spanning corporate transactions, real estate, and estate planning, they'd built a solid reputation over 40 years of service.
Like many law firms, they prided themselves on being forward-thinking—adopting technology to serve clients more efficiently.
The Incident
A junior associate was working on a complex motion under a tight deadline. Facing pressure to deliver, she turned to ChatGPT to help research relevant case law and draft portions of the brief.
The AI was impressively helpful. It produced well-written legal arguments supported by what appeared to be relevant case citations. The associate incorporated the AI's suggestions into the brief, which went to her supervising partner for review.
The partner, also under time pressure, focused on the legal arguments rather than individually verifying each citation. The brief was filed.
The Discovery
Opposing counsel, as experienced attorneys do, checked the citations. They found a problem—three of the cases didn't exist.
The AI had "hallucinated" case citations: generating fictional case names, courts, and dates that looked authentic but were completely fabricated.
Opposing counsel filed a motion for sanctions.
The Court Response
The judge was not amused:
"Counsel's submission of fabricated case citations represents a fundamental breach of the obligations attorneys owe to this court."
Sanctions imposed: $15,000
The ruling was published. Legal media picked up the story.
The Cascading Consequences
State Bar Investigation
The state bar opened an ethics investigation into the supervising partner. Questions centered on:
- Duty of competence in supervising AI-assisted work
- Candor to the tribunal
- Responsibility for associate errors
Legal defense costs: $45,000
The investigation ultimately resulted in a private reprimand, but the process consumed months of the partner's time and attention.
The Client Lawsuit
The client whose matter was damaged by the sanctions and publicity retained new counsel—and sued Harrison & Cole for malpractice.
Settlement: $125,000
Insurance Impact
At renewal, the firm's malpractice insurance carrier:
- Questioned their AI usage policies
- Required documentation of new controls
- Increased premiums by 85%
Reputation Damage
The story appeared in:
- American Lawyer
- Law360
- State bar journal
- Local legal press
Prospective clients researching the firm found the coverage immediately.
Partner Departure
Six months after the incident, the supervising partner left the firm. The combination of the bar investigation, client relationship damage, and internal stress proved too much.
Total Impact
| Category | Cost |
|---|---|
| Court sanctions | $15,000 |
| Bar defense costs | $45,000 |
| Malpractice settlement | $125,000 |
| Insurance premium increase (Year 1) | $35,000 |
| Lost partner (recruiting replacement) | ~$50,000 |
| Direct Costs | $270,000+ |
The reputational damage and lost business opportunities are impossible to quantify but certainly substantial.
Root Cause Analysis
How did a 40-year-old firm end up here?
No AI Policy
The firm had no guidance on AI usage. Associates made individual decisions about whether and how to use AI tools.
No Verification Protocol
There was no required process for verifying AI-generated content. The assumption was that attorneys would apply professional judgment—but there was no framework for what that meant with AI.
Training Gap
Neither the associate nor the partner understood AI hallucination as a phenomenon. They didn't know to look for fabricated citations because they didn't know AI could fabricate so convincingly.
Time Pressure Culture
Legal work often operates under crushing deadlines. That pressure created conditions where shortcuts seemed necessary.
The Remediation
Harrison & Cole implemented comprehensive changes:
1. AI Verification Protocol
All AI-generated content must be verified before submission:
- Every citation must be checked in an authoritative database
- AI-generated legal arguments must be independently confirmed
- A verification checklist must accompany every court filing
2. Approved Tool List
The firm identified AI tools approved for specific uses:
- Legal research: Approved tools with citation verification
- Document drafting: Enterprise tools with audit trails
- General productivity: Specific approved services only
3. Mandatory Training
All attorneys and staff completed training covering:
- AI capabilities and limitations
- Hallucination risks
- Verification requirements
- Professional responsibility implications
4. Supervision Requirements
Partners must specifically attest to AI verification in any submission where AI assisted with research or drafting.
The Managing Partner's Reflection
"We prided ourselves on being innovative. But innovation without governance nearly destroyed a 40-year-old firm. Every law firm needs an AI policy—not next year, today."
Lessons for Law Firms
1. AI Hallucination is Real and Dangerous
Large language models can generate convincing but entirely fictional citations, facts, and analysis. This isn't a bug—it's a fundamental characteristic of how these systems work.
2. Professional Responsibility Extends to AI
The duty of competence includes understanding the tools you use. The duty of candor means verifying what you submit to courts.
3. Supervision Models Must Adapt
Traditional supervision assumes human sources. AI outputs require different verification approaches.
4. The Risk Isn't the Technology—It's the Governance Gap
AI can be valuable for legal work. The danger is using it without proper frameworks for verification and oversight.
5. Post-Incident Costs Dwarf Prevention Costs
A comprehensive AI governance program might cost a law firm $5,000-15,000 to implement. This incident cost $270,000+.
What Every Law Firm Needs
Minimum AI Governance for Legal Practice:
- Written AI Policy — What tools are approved, for what uses
- Verification Protocol — Required steps before relying on AI output
- Training Program — All attorneys understand AI limitations
- Supervision Framework — Clear responsibilities for AI-assisted work
- Audit Trail — Documentation of AI usage in client matters
Is Your Firm Protected?
If your attorneys are using AI—and statistically, they are—do you have the governance framework to prevent an incident like this?
Take our free AI Risk Assessment to evaluate your firm's exposure.
This case study is a composite based on real-world incidents, including the widely-publicized Mata v. Avianca case and similar situations. Details have been modified to illustrate common patterns while protecting confidentiality.