ChatGPT Erroneously Alleged Mayor Served Prison Time for Bribery

ChatGPT erroneously alleged regional Australian mayor Brian Hood served time in prison for bribery. Mayor Hood is considering legal action against ChatGPT's makers for alleging a foreign bribery scandal involving a subsidiary of the Reserve Bank of Australia in the early 2000s.

Risk classification

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

The AI system generated specific false and defamatory statements about an individual, leading to reputational harm.

Additional risk subdomains

• 7.3 Lack of capability or robustness: The incident stems from the AI model's technical limitation of hallucinating plausible but incorrect information.

Causal factors

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

The defamatory statements were generated by the ChatGPT AI system post-deployment as an unintended hallucination during text generation.

EU AI Act risk tier

• Risk tier: 3 Limited Risk

Risk Level 3: Limited Risk. The system is a chatbot (ChatGPT) which poses a moderate risk and is subject to transparency obligations to ensure users are aware they are interacting with an AI.

AI system and alleged parties

• AI system: Bing, ChatGPT (Microsoft, OpenAI)
• AI purpose: Chatbot; Writing Assistant
• Behaviour type: Assistant
• Alleged developer: OpenAI
• Alleged deployer: OpenAI
• Alleged harmed parties: Brian Hood

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

Epistemic

Reported: Yes, the report explicitly describes ChatGPT generating false and fabricated information about Brian Hood's past.

Directly caused: ChatGPT fabricated a false narrative that Brian Hood was convicted of bribery and sentenced to prison, completely reversing his actual role as a whistleblower.

Indirectly caused: N/A

Inferred additional harm: N/A

People affected

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

Potential causes

Management

• Insufficient Disclaimers: Generic warnings do not mitigate severe, reputational harms generated.
• Slow Response to Complaints: Delayed response to legal letters demanding retraction of false claims.

Technology

• LLM Hallucination: Model trained to produce plausible text rather than factual accuracy.
• Opaque Neural Network: Difficult to trace or surgically remove specific false facts from weights.
• Lack of Citations: Output lacked footnotes or sources, creating false sense of accuracy.

Data Inputs

• Associative Training Data: Model associates Hood's name with bribery due to co-occurrence in texts.

Human Factors

• User Overreliance: Users trust chatbot outputs for factual research without verification.

Process and Methods

• Inadequate Fact-Checking: No real-time verification process to prevent false claims about people.
• Ineffective Error Correction: No mechanism to easily locate and delete specific false facts in network.

Regulatory Environment

• Lack of AI Regulation: No clear legal standards governing defamatory outputs from AI systems.

Information quality

• Classification confidence: High
• Reason for confidence: The reports provide clear, consistent, and detailed accounts of the false claims made by ChatGPT about Brian Hood, his actual role as a whistleblower, and his subsequent legal response. There is no conflicting information regarding the core facts of the incident.

An Australian mayor threatened legal action against OpenAI after ChatGPT falsely accused him of bribery and prison time, reversing his actual role as a whistleblower. This incident highlights the national security and legal challenges of AI-generated misinformation affecting public officials, though it poses minimal immediate threat to state sovereignty or critical infrastructure.

• Overall national security impact: Minor
• Response level: Moderate
• Scope: Single nation
• Primary target: Australia
• Alleged perpetrator: Unknown

Threat characteristics

• Imminence: Long-term. The incident represents a long-term policy and legal challenge regarding AI accountability rather than an active crisis.
• Autonomy: Human-controlled. The AI model generates text autonomously based on prompts, but lacks operational agency or independent decision-making capabilities.
• Novelty: Evolved capability. While AI hallucinations are well-known, this incident represents an evolved capability where LLMs generate highly specific, plausible, and damaging falsehoods about public figures.

Impact by dimension

• Physical security: Negligible. No physical systems, critical infrastructure, or human safety elements were impacted or threatened by this incident.
• Information security: Negligible. The incident involves an unintentional AI hallucination rather than a coordinated foreign disinformation campaign or intelligence compromise.
• Sovereignty: Minor. An elected local official was affected by false claims, but there was no disruption to core government functions, sovereignty, or electoral systems.
• Economic security: Negligible. The event represents a localized legal dispute with no impact on strategic industries, financial systems, or technological competitive advantage.
• Societal stability: Minor. The incident caused individual reputational and psychological harm to a single public official, but did not threaten broader social cohesion or civil liberties.