3M-Retained Expert Reportedly Submitted Largely ChatGPT-Generated Analysis Seeking to Exonerate 3M as Expert Evidence in Explosion Lawsuit

3M-retained expert Josh Autenrieth reportedly submitted an expert report substantially generated with ChatGPT after prompting it to show 3M was "0% at fault" for a 2020 Houston industrial explosion that killed three people and destroyed roughly 200 homes. Plaintiffs were thereby presented with purported expert evidence substantially drafted toward a requested conclusion by an AI system, prompting extensive discovery and trial scrutiny.

Risk classification

• Primary risk domain: 5 Human-Computer Interaction
• Primary risk subdomain: 5.1 Overreliance and unsafe use

The expert witness exhibited extreme overreliance on ChatGPT, using it to form 85% to 90% of his professional opinion in a critical legal and safety-related lawsuit.

Additional risk subdomains

• 4.3 Fraud, scams, and targeted manipulation: The expert witness used the AI system to generate biased arguments to intentionally manipulate the court's perception of liability for his client's benefit.
• 3.1 False or misleading information: The AI was prompted to generate misleading and biased expert testimony to falsely claim 3M was 0% responsible for the explosion.

Causal factors

• Entity: Human
• Intent: Intentional
• Timing: Post-deployment

The incident was caused by the human expert witness's intentional decision to use ChatGPT to generate a biased expert report to influence a lawsuit.

EU AI Act risk tier

• Risk tier: 3 Limited Risk

Limited Risk: The system used is ChatGPT, a general-purpose chatbot. Under the EU AI Act, chatbots are classified as limited risk and are subject to transparency obligations to ensure users know they are interacting with AI.

AI system and alleged parties

• AI system: ChatGPT (OpenAI)
• AI purpose: Writing Assistant; Technical Text Generation
• Behaviour type: Assistant
• Alleged developer: OpenAI, Large language model developers, Chatbot developers
• Alleged deployer: Josh Autenrieth, Expert witnesses, engineers
• Alleged harmed parties: Plaintiffs in Watson Grinding explosion litigation, Legal system, Judicial integrity, Epistemic integrity, Courts

Harm severity

Severity ratings are based on this scale.

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 Minor, indirect Negligible
• 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 Negligible, indirect Negligible
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Financial

Reported: The report describes 1 incident causing a financial loss of $90,000 in wasted expert fees paid by 3M.

Directly caused: 3M paid Knighthawk Engineering approximately $90,000 for an expert report that was 85% to 90% generated by ChatGPT, which was ultimately exposed and undermined their defense.

Inferred additional harm: 3M likely incurred additional unquantified legal costs due to the trial delay and the embarrassment of the AI exposure during depositions.

People affected

• Occurrences reported: 1
• People reportedly exposed: 50

Potential causes

Management

• Lack of Vendor Oversight: 3M failed to monitor Knighthawk Engineering's report preparation methods.
• Inadequate Quality Assurance: No internal review detected that the report was largely machine-generated.

Technology

• Generative AI Limitations: ChatGPT outputs structure and content that may lack rigorous expert validation.
• Lack of Input Bias Filtering: AI model accepted highly biased prompts without correcting or flagging them.

Data Inputs

• Biased Prompting: Prompts instructed ChatGPT to prove 3M was 0% responsible for the explosion.
• Direct Upload of Court Records: Feeding hundreds of court records directly into AI to generate legal arguments.

Human Factors

• Overreliance on AI Output: Expert relied on ChatGPT to write 85% to 90% of his technical expert report.
• Lack of AI Privacy Awareness: Expert exposed sensitive prompts publicly through ChatGPT links.
• Compromised Objectivity: Expert directed AI to create a biased defense instead of objective analysis.

Process and Methods

• No Verification of AI Drafts: Failing to independently verify technical claims generated by ChatGPT.
• Insecure Tool Usage: Conducting sensitive litigation work on public AI platforms.

Regulatory Environment

• Unclear Rules on AI in Court: Lack of clear regulations regarding AI use in expert witness testimonies.

Information quality

• Classification confidence: High
• Reason for confidence: The report provides highly detailed, factual information about the incident, including specific quotes from the expert's prompts, details of the court proceedings, and the financial figures involved.

An expert witness in a major civil lawsuit used ChatGPT to generate 85-90% of an expert report, prompting the AI with highly biased instructions to absolve his client of liability. While exposed during discovery and resolved via standard judicial processes, the incident highlights an evolved threat to judicial integrity through AI-generated biased testimony in court proceedings.

• Overall national security impact: Minor
• Scope: Single nation
• Primary target: United States
• Alleged perpetrator: Autenrieth of Knighthawk Engineering

Threat characteristics

• Imminence: Long-term. Represents an ongoing strategic concern regarding the integrity of legal and judicial processes rather than an active, immediate crisis.
• Autonomy: Human-controlled. The AI system acted as a writing assistant, generating content based on direct human prompts, with the human expert retaining control over submission.
• Novelty: Evolved capability. Demonstrates an evolved capability of using generative AI to systematically draft biased technical testimony, establishing a clear legal precedent for AI discovery.

Impact by dimension

• Physical security: Negligible. The AI was used purely for drafting legal reports post-incident, with no direct impact on physical systems, targeting, or critical infrastructure operations.
• Information security: Negligible. No foreign state-sponsored information warfare, intelligence compromise, or mass-scale disinformation campaigns occurred in this localized civil lawsuit.
• Sovereignty: Minor. Represents a minor threat to judicial integrity by introducing highly biased, AI-generated expert testimony into a domestic court proceeding, though resolved through standard discovery.
• Economic security: Negligible. Financial impacts are limited to standard civil litigation damages and localized corporate legal fees, posing no threat to national economic or technological security.
• Societal stability: Negligible. No population-scale societal manipulation, civil liberties violations, or threats to social cohesion were present in this corporate legal dispute.