AI Risk & Hallucination Cases
A regularly updated digest of documented AI-related risk incidents in legal practice — primarily court cases involving AI-generated citation errors, sanctions, and judicial orders, but also significant malpractice or ethics matters. Each entry summarizes the incident, identifies the AI tool involved where known, notes the court, jurisdiction, and outcome (including any sanctions), and links to the primary court record or credible reporting. This group serves risk officers, malpractice insurers, and attorneys who need to understand real-world failure modes. It is distinct from regulatory-tracker (which covers rules and obligations) and from news-and-analysis (which covers broader market events). Entries must carry explicit disclaimers that summaries are informational and not legal analysis of the cases.
Risk incident entries
AI Citation Hallucination Sanctions in Federal Courts: Case Patterns, Penalties, and Enforcement Principles (2023–Q1 2026)
A structured reference covering the full arc of U.S. federal court AI citation sanctions — from the 2023 watershed through Q1 2026's record $145,000 quarter — documenting verified case counts, sanction amounts, the legal authority bases courts apply, and the four enforcement principles that now give litigating attorneys a calculable risk model.
- AI hallucination / fabricated citationMiddle District of Florida
AI Hallucinations and Attorney Ethics: Which Professional Responsibility Rules Are Triggered and How Sanctions Have Escalated
This risk-digest article provides practicing attorneys and in-house counsel with a rule-by-rule analysis of the professional responsibility obligations triggered when AI hallucinations reach court filings, supported by an annotated case table of sanctions from 2023 through mid-2026 and a practical risk-calibration framework.
- Outcome
- 1-year suspension from federal bar with reinstatement conditions
- Incident date
- 2024-03
- AI hallucination / fabricated citationU.S. federal and state courts (multiple jurisdictions)
AI Hallucinations in Legal Practice: The Sanctions Trajectory and the Verification Discipline Every Lawyer Must Adopt
This article traces the enforcement trajectory of AI-generated hallucinations in legal filings from the 2023 Mata v. Avianca $5,000 sanction to the record $145,000 in Q1 2026 penalties, and argues that the profession's failure to operationalize verification discipline — not AI unreliability alone — is the root problem. It provides litigators, ethics partners, and risk officers with the data, frameworks, and protocols needed to navigate the new enforcement reality.
- Outcome
- $145,000 in Q1 2026 sanctions; $109,700 Oregon aggregate; $30,000 per attorney federal appellate fine
- AI tool involved
- ChatGPT
- Incident date
- 2026-03-31
- court orderN.D. Cal. / S.D.N.Y. (federal)
AI Litigation by the Numbers: Case Volume, Venue Concentration, and Defendant Exposure in 2025–2026
A data-driven baseline for litigation risk officers and in-house counsel: AI-related federal cases surged from 7 in 2022 to 94 in 2025, concentrated in the N.D. Cal. and S.D.N.Y., with OpenAI named in over 40% of all tracked filings. This article provides the quantitative foundation for sizing AI litigation risk, budgeting for defense costs, and evaluating insurance exposure.
- Outcome
- Data-driven baseline for risk assessment; 94 cases filed in 2025
- AI tool involved
- OpenAI
- Incident date
- 2026-06-18
- malpractice claimCalifornia, United States
Legal Implications of AI Universal Basic Income Proposals
No federal AI-linked universal basic income bill has passed as of mid-2026, but pilot programs, state causation-based payment schemes, and sovereign wealth fund proposals are rapidly building a legislative architecture that legal professionals must understand before compliance obligations crystallize.
- Outcome
- No enacted federal law as of Q3 2026
- Incident date
- 2026-07-20
- business model disruptionUnited States (general)
AI Will Not Replace Lawyers — But It Will Replace Law Firms That Refuse to Adapt
A strategic analysis for law firm leaders arguing that the real disruption from AI is not job displacement but structural change to firm economics, client expectations, and the billable hour model. Firms that treat AI as a checkbox feature risk losing clients to in-house teams and more innovative competitors.
- Outcome
- No sanctions or court action; strategic analysis of market shift
- AI tool involved
- general AI tools
- Incident date
- 2026-06-19
- court orderSouthern District of New York
Claude outages in 2026 create compound ethics and privilege risk
Claude's 2026 reliability record—50+ incidents, two-tier uptime, and an 84% actual uptime claim—creates exposure patterns most law firms have not modeled: a single outage during deadline week can simultaneously trigger privilege waiver under United States v. Heppner, violate ABA Formal Opinion 512 competence requirements, and breach Texas Opinion 705's independent verification mandate.
- Outcome
- Privilege denied; conviction on all counts
- AI tool involved
- Claude
- Incident date
- 2026-02-17
- AI hallucination / fabricated citationN/A (business model analysis, not a court case)
Crosby's Per-Document Pricing: A Concrete Case Study in Inverting the Billable Hour
This article examines Crosby, an AI-native law firm that has abandoned the billable hour for fixed per-document pricing ($250–$1,000 per contract). For law firm partners, legal ops leaders, and industry analysts, it analyzes how this model structurally aligns firm incentives with client goals and what it signals for the future of legal services delivery.
- Outcome
- N/A (business model analysis, not a court case)
- AI tool involved
- Crosby
- Incident date
- 2026-03-31
- AI hallucination / fabricated citationSan Francisco Superior Court, California
Divine prophecy suicide claim pushes OpenAI liability boundaries
The Christian Faith Madison wrongful death lawsuit against OpenAI introduces an unprecedented product-liability theory: that ChatGPT's sycophantic architecture fabricated a divine prophecy narrative driving the user to suicide. This case analysis examines the allegations, contrasts them with prior suicide-coaching suits, and identifies unresolved legal questions for the evolving liability landscape.
- Outcome
- Lawsuit filed; awaiting response
- AI tool involved
- GPT-4o
- Incident date
- 2025-06-08
- cyber incidentUnited States
What the Fairlife ransomware attack means for dairy supply contracts
When a ransomware attack shuts down a major dairy producer, supply contracts face disruptions that standard force majeure clauses may not cover. This analysis examines how the Fairlife incident tests force majeure, UCC allocation rules, and what counsel should do to protect their clients' positions.
- Outcome
- Legal analysis published; no court ruling
- Incident date
- 2026-07-16
- AI hallucination / fabricated citationFederal courts (U.S.)
Over 60% of Federal Judges Use AI in Chambers, But Training and Policies Lag Behind
A 2026 random-sample survey of 502 federal judges reveals that more than 60% use at least one AI tool in chambers, yet nearly half received no AI training from court administration and one in four has no official AI policy. This data-driven risk-digest article analyzes the survey findings, the tool preferences of judges, the policy patchwork, and the implications for litigants and attorneys navigating an inconsistent judicial AI landscape.
- Outcome
- Survey data; Grassley investigation; sanctions including disqualification
- AI tool involved
- Westlaw AI-Assisted Research
- Incident date
- 2026-03-30
- sanctionsU.S. federal and state courts
Why Ford's 2016-2019 recalls matter for product liability risk
A source-cited digest of Ford's recall incidents, litigation outcomes, and regulatory actions from 2016 to 2019. It reveals how delayed defect disclosure produced cascading liability across multiple defect categories.
- Outcome
- Regulatory consent order with $165M civil penalty, settlements, ongoing litigation
- Incident date
- 2018-07
- court orderD. Del. (Thomson Reuters v. Ross); N.D. Cal. (Bartz v. Anthropic, Kadrey v. Meta)
The Great Fair Use Divide: Three AI Training Rulings, Three Different Outcomes, and What They Mean for 2026
For in-house IP counsel and AI product attorneys: a comparative analysis of the three pivotal 2025 fair use rulings — Thomson Reuters v. Ross, Bartz v. Anthropic, and Kadrey v. Meta — and how the fractured landscape shapes litigation risk, licensing strategy, and compliance obligations in 2026.
- Outcome
- Thomson Reuters: fair use rejected, appeal pending; Bartz: training fair use, piracy infringing, settled for $1.5B; Kadrey: training fair use, judge invites stronger market-harm evidence
- AI tool involved
- Claude, LLaMA
- Incident date
- 2025-06-25
- AI hallucination / fabricated citationMultiple jurisdictions (global)
Hallucinations, Ethical Walls, and the Verification Paradox: The Real Risk Profile of Harvey AI for Law Firms
A risk-focused analysis for legal leaders evaluating Harvey AI deployment, examining the gap between vendor reliability claims and the operational verification burden imposed by ABA Formal Opinion 512, Harvey-specific hallucination incidents, and ethical wall failure modes for AI agents.
- Outcome
- Risk analysis; no single outcome — community-documented fake LexisNexis citation (April 2026)
- AI tool involved
- Harvey
- Incident date
- 2026-04-01
- court orderCalifornia
Hyundai Tucson software recalls expose a new liability category
Using Hyundai's two 2026 software recalls as paired case studies, this article examines how automotive software defects create structurally different product liability exposure from traditional hardware recalls—and what those differences mean for in-house counsel and defense attorneys navigating recall documentation and class-action risk.
- Outcome
- Recall issued; free software update; class action filed
- Incident date
- 2026-05
Review tool profiles for products mentioned in incidents
Tool profiles include data privacy posture and accuracy benchmark data relevant to risk assessment.
Browse tool profiles →Check regulatory obligations related to these incidents
The regulatory tracker covers the court orders, bar opinions, and legislation relevant to these failure modes.
Open regulatory tracker →