Canadian MLA Bill Oliver Reportedly Read Unremoved LLM Instructions During New Brunswick Legislative Speech

During a June 9, 2026 speech in the Legislative Assembly of New Brunswick, Canada, MLA Bill Oliver read aloud two passages that appeared to be unremoved instructions from an LLM-assisted draft. The remarks were preserved in the Assembly's official recording.

Risk classification

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

The primary harm arose because a politician and multiple lawyers placed inappropriate trust and reliance on AI systems to generate professional text without verifying the outputs.

Additional risk subdomains

• 3.1 False or misleading information: The AI systems used by the lawyers generated fabricated case law and quotes, which were then introduced into court filings as factual.

Causal factors

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

The incidents were caused by human decisions to use AI-generated text without proper review, leading to unintended public embarrassment and legal sanctions.

EU AI Act risk tier

• Risk tier: 3 Limited Risk

Limited Risk: The report describes the use of chatbots and AI-generated text, which fall under limited risk requiring transparency obligations so users know they are interacting with AI or using AI-generated content.

AI system and alleged parties

• AI purpose: Writing Assistant; Technical Text Generation
• Behaviour type: Assistant
• Alleged developer: Large language model developers, Chatbot developers
• Alleged deployer: Political officeholders, Members of legislative bodies, Bill Oliver
• Alleged harmed parties: Public trust, Legislative proceedings, Legislative Assembly of New Brunswick, General public of New Brunswick, General public of Canada, General public

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 not assessed, indirect not assessed
• Infrastructure: direct not assessed, indirect not assessed
• Property: direct not assessed, indirect not assessed
• Financial: direct Negligible, indirect not assessed
• Environmental: direct not assessed, indirect not assessed
• Malicious content: direct not assessed, indirect not assessed
• Differential treatment: direct not assessed, indirect not assessed
• Civil rights: direct not assessed, indirect not assessed
• Democracy: direct Negligible, indirect not assessed
• Privacy: direct not assessed, indirect not assessed
• Psychological: direct not assessed, indirect not assessed
• Epistemic: direct Minor, indirect not assessed
• Child sexual exploitation and abuse: direct not assessed, indirect not assessed

Epistemic

Reported: Yes, the report describes lawyers filing court documents riddled with AI hallucinations, including fabricated case law and quotes.

Directly caused: Lawyers filed court documents containing fabricated case law and quotes generated by AI models.

Inferred additional harm: The introduction of fabricated legal precedents into court records threatens the integrity of the judicial system and the accuracy of legal proceedings.

People affected

• Occurrences reported: 3
• People reportedly harmed: 7
• People reportedly exposed: 7

Potential causes

Management

• Low Accountability Standards: Lack of strict professional standards for politicians compared to other fields.

Technology

• Chatbot Output of Meta-text: AI chatbot included conversational meta-text in the generated speech text.
• Convincing Mimicry of Speech: AI sounds passable enough to tempt users to use it without close review.

Human Factors

• Inattentive Speech Delivery: Lawmaker read the text verbatim without noticing the AI prompt instruction.
• Over-reliance on AI Tools: Outsourcing speechwriting to save effort instead of drafting original text.
• Lack of Care for Audience: Prioritizing convenience over genuine communication with the public.

Process and Methods

• Lack of Pre-speech Review: No process to proofread or verify speech content before public delivery.
• Direct Copy-Pasting: Copying raw chatbot output directly into official speech notes.

Regulatory Environment

• Lack of Rules for Politicians: No formal regulations or codes of conduct governing AI use in legislatures.

Information quality

• Classification confidence: High
• Reason for confidence: The report clearly details the incidents of a politician reading an AI prompt and lawyers being sanctioned for AI hallucinations. The roles of the AI systems as text generators and the human errors in overrelying on them are explicitly described.
• Ambiguities identified: The specific AI models used by the politician and the lawyers are not named.
• Alternative interpretations: The lawyers' incidents could be classified primarily under lack of capability (7.3) due to hallucinations, but the overarching theme is human overreliance on these tools.

This incident involved a Canadian lawmaker and several US lawyers who failed to verify AI-generated text, leading to public embarrassment and court sanctions. The national security impact is minor, primarily affecting the integrity of legislative and judicial processes through professional overreliance on AI, rather than malicious targeting or systemic threat.

• Overall national security impact: Minor
• Scope: Multiple nations
• Primary target: No clear primary
• Other affected: Canada, United States
• Alleged perpetrator: Unknown

Threat characteristics

• Imminence: Long-term. Represents an ongoing strategic concern regarding professional overreliance on AI tools rather than an active crisis.
• Autonomy: Human-controlled. The AI systems acted as writing assistants, with humans retaining final control and execution authority, despite failing to verify the outputs.
• Novelty: Established threat. Similar incidents of AI hallucinations in legal filings and professional negligence have been documented previously.

Impact by dimension

• Physical security: Negligible. No physical systems, critical infrastructure, or kinetic targeting capabilities were affected by this incident.
• Information security: Negligible. No intelligence compromise or systematic, state-sponsored information warfare operations were identified in this incident.
• Sovereignty: Minor. Minor impact on government and judicial functions. A Canadian lawmaker and US court officers introduced unverified AI-generated content into official proceedings, resulting in minor procedural disruptions and public embarrassment.
• Economic security: Negligible. No strategic technology theft, critical supply chain vulnerabilities, or threats to economic stability were reported.
• Societal stability: Negligible. No threats to societal stability, civil liberties, or population safety were reported.