D.C. Court of Appeals Struck Deutsche Bank National Trust Company Brief After Attorney Used Google Generative AI Search Tool and Filed Four Nonexistent Case Citations

In Douglas v. Deutsche Bank National Trust Co., attorney Loishirl W. Hall acknowledged using Google's generative AI search tool to locate case authority and failing to verify citations before an appellate brief was filed. The D.C. Court of Appeals found four cited cases did not exist, said the filing misdirected the court and burdened judicial resources, struck Deutsche Bank's brief, and referred the matter for disciplinary review.

Risk classification

• Primary risk domain: 3 Misinformation
• Primary risk subdomain: 3.1 False or misleading information

The AI system generated fabricated legal citations (hallucinations) which were presented as real authority in a court brief, leading to epistemic harm and legal penalties.

Additional risk subdomains

• 5.1 Overreliance and unsafe use: The attorney overrelied on the AI tool for legal research and failed to verify the citations before submitting them to the court.
• 7.3 Lack of capability or robustness: The AI system failed to perform reliably by hallucinating nonexistent legal cases.

Causal factors

• Entity: AI
• Intent: Unintentional
• Timing: Post-deployment

The incident was initiated by the AI system generating fabricated legal citations (hallucinations) during post-deployment use, which was an unexpected and unintentional outcome of the legal research query.

EU AI Act risk tier

• Risk tier: 3 Limited Risk

Limited Risk: The Google AI program used is a generative AI tool capable of generating text and citations, which falls under the category of AI-generated content requiring transparency and verification.

AI system and alleged parties

• AI system: Google generative AI search tool (Google)
• AI purpose: Writing Assistant; Content Search
• Behaviour type: Assistant
• Alleged developer: Google
• Alleged deployer: Loishirl W. Hall, Legal professionals, Lawyers
• Alleged harmed parties: Parties to legal proceedings, Legal clients, Judicial integrity, Epistemic integrity, District of Columbia Court of Appeals, Courts

Harm severity

Highest direct severity in any category: Minor. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

• Physical: direct Negligible, indirect Negligible
• Infrastructure: direct Negligible, indirect Negligible
• Property: direct Negligible, indirect Negligible
• Financial: direct Negligible, indirect Minor
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Negligible, indirect Negligible
• Differential treatment: direct Negligible, indirect Negligible
• Civil rights: direct Negligible, indirect Negligible
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Negligible, indirect Negligible
• Psychological: direct Negligible, indirect Negligible
• Epistemic: direct Minor, indirect Minor
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Financial

Reported: The report does not explicitly quantify financial losses, but notes the client was penalized by having their brief struck, which likely carries indirect financial consequences.

Directly caused: N/A

Indirectly caused: The client, Deutsche Bank, likely incurred indirect financial losses due to their brief being struck in a foreclosure action, requiring additional legal expenses or risking the loss of the foreclosure lawsuit.

Inferred additional harm: The law firm and the client likely faced thousands of dollars in wasted billable hours and potential legal liabilities arising from the struck brief.

Epistemic

Reported: Yes, the report explicitly describes the fabrication of legal citations which misdirected the court and sowed confusion.

Directly caused: The AI system generated four nonexistent case citations (Abadie v. District of Columbia, Consumer Co-op Bank, Osborne v. District of Columbia, and Woods v. United States) which were submitted to the court as legitimate authority.

Indirectly caused: The court had to devote substantial time and resources to verify the fake citations rather than addressing the merits of the appeal, wasting judicial resources.

Inferred additional harm: N/A

People affected

• Occurrences reported: 1
• People reportedly harmed: 1
• People reportedly exposed: 5

Potential causes

Management

• Ineffective Policy Enforcement: Firm policy banning AI drafting was not actively enforced or monitored.
• Deficient Supervisory Oversight: The firm failed to actively supervise the work of a novice associate.

Technology

• AI Propensity to Hallucinate: Generative AI tools fabricate nonexistent legal citations as real authority.
• AI Susceptibility to Inaccuracies: AI models make factual errors and mischaracterize holdings of real cases.

Human Factors

• Blind Reliance on AI Outputs: Attorney failed to independently verify the AI-generated case citations.
• Inexperience of Submitting Attorney: The attorney had never drafted or filed an appellate brief prior to this case.
• Lack of Understanding of AI Risks: Failure to appreciate the technology's propensity to invent legal authority.

Process and Methods

• Inadequate Brief Review Process: Other signing attorneys failed to vet the brief's citations before filing.
• Lack of Pre-filing Verification: No process was in place to verify citations obtained via AI search tools.

Regulatory Environment

• Ambiguous Appellate Sanction Rules: Local rules lack clear provisions for directly sanctioning negligent attorneys.

Information quality

• Classification confidence: High
• Reason for confidence: The court order provides highly detailed, authoritative, and verified factual accounts of the incident, including the specific fake citations, the attorney's admission, and the court's response.
• Ambiguities identified: The specific Google AI model used by the attorney is not named, referred to only as a Google artificial intelligence program.
• Alternative interpretations: None. The facts of the AI hallucination and subsequent court filing are clear and uncontested.

An attorney used a Google AI tool that generated fabricated case citations in a court filing. While it disrupted local judicial proceedings and led to professional discipline, it has negligible national security implications.

• Overall national security impact: Negligible
• Response level: Minor
• Scope: Single nation
• Primary target: United States
• Alleged perpetrator: Loishirl Hall

Threat characteristics

• Imminence: Long-term. This represents a minor, isolated professional conduct issue rather than an active national security threat or crisis.
• Autonomy: Human-controlled. The AI served as an assistant tool for drafting, with the human attorney retaining final control and responsibility for the filing.
• Novelty: Established threat. AI-generated legal hallucinations are an established phenomenon with multiple documented precedents in the legal system.

Impact by dimension

• Physical security: Negligible. No physical infrastructure, weapon systems, or human safety concerns were involved in this legal filing incident.
• Information security: Negligible. The incident involves localized legal research errors rather than systematic disinformation, state-sponsored influence, or intelligence compromise.
• Sovereignty: Negligible. The disruption was limited to a single local appellate court proceeding and did not threaten core government functions or state authority.
• Economic security: Negligible. The financial consequences were strictly private, affecting only the client bank and the law firm, with no systemic economic impact.
• Societal stability: Negligible. The event did not involve mass surveillance, civil liberties violations, or threats to societal stability.