Bill Oliver reads AI prompts aloud in New Brunswick legislature speech

A New Brunswick politician has sparked international attention and renewed debate regarding political transparency after accidentally reading artificial intelligence prompt instructions aloud during a legislative speech. Bill Oliver, a Progressive Conservative member representing Kings Centre, delivered the address on June 9, touching off a wave of scrutiny over the growing reliance on automated writing tools in public office.

The gaffe occurred while Oliver spoke during a legislative discussion centered on whether the province should establish an official energy advocate, as reported by Global News. Rather than delivering a purely human-composed address, Oliver read directly from speaking notes generated by an AI chatbot. Captured on the official legislative record and subsequently circulated widely across platforms like Reddit and Instagram, the speech featured jarring meta-instructions, including an audible pause where Oliver stated,

"Here’s a more natural, flowing version of that section that reads like a legislative speech rather than a series of short points."

Media additions

Bill Oliver, MLA for Kings Centre, via Global News

According to coverage by the CBC, Oliver later suffered two additional slips during the address, reading phrases such as Here's a more developed and flowing version of the same section and This section adds roughly … excuse me. The incident quickly went global, drawing headlines from outlets including Al Jazeera, Euronews, India Today, and Kenya’s Tuko News.

In a formal statement released on a Wednesday, Oliver acknowledged the error, noting that he failed to properly review the text before taking the floor.

"When printing the final version of my speech, AI prompts were not removed, which were spoken by me and has caused much concern of many individuals,"

Bill Oliver, MLA for Kings Centre, via Global News and CBC

Oliver said, while maintaining that the underlying sentiments of the speech were entirely his own. Glen Savoie, the interim PC leader, defended Oliver, telling reporters that the MLA had taken responsibility for a human mistake and noting that the caucus is working on creating an artificial intelligence policy for its members.

The fallout has triggered sharp divisions across the provincial political landscape. Green Party representative Megan Mitton, who noted that she and fellow Green MLA David Coon immediately recognized the AI phrasing on the day of the delivery, argued that the lengthy address stalled the legislature's core work. Mitton asserted that lawmakers are elected to voice the concerns of their constituents rather than rely on automated text generators. While the Liberal caucus indicated it maintains no formal guidelines on AI usage beyond existing obligations regarding ethics and professional conduct, the Green caucus adopted an internal policy restricting AI speechwriting in March 2026.

Independent analysis underscores how widespread the phenomenon has become. Jon Gillham, CEO of Originality AI, revealed that an analysis of approximately 8,000 English-language speeches delivered in the New Brunswick legislature between 2022 and 2026 showed a dramatic upward trajectory in automated text usage. According to Originality AI's findings, 3.3 per cent of speeches likely contained AI-generated text in 2022, rising to 21 per cent in the previous year, and reaching almost 23 per cent of speeches in 2026.

Meanwhile, academic experts have raised alarms over the implications of unverified AI outputs in public policy. Mahtab Sarvmaili, director of the Responsible AI Research Lab at the University of New Brunswick, warned that large language models rely on pre-existing data and require rigorous fact-checking to prevent factual errors. Mario Levesque, a political science professor at Mount Allison University, echoed those concerns, emphasizing that while AI can serve as a functional tool, public officials must verify all underlying facts.

Broader public sentiment strongly aligns with these concerns. Data from the Pew Research Center indicates that 71 per cent of U.S. Adults would view a political candidate less favourably upon discovering that AI helped write a speech they otherwise liked. Public sector adoption of generative tools is not restricted to Canadian legislatures; investigative reporting from KNKX Public Radio demonstrated that municipal officials in Washington state frequently utilize systems like ChatGPT for mayoral correspondence, grant applications, and constituent replies, often operating without mandatory public disclosures or clear oversight guardrails.

As legislative bodies grapple with the rapid acceleration of generative technology, watchdogs emphasize that transparency remains paramount to safeguarding public trust in democratic institutions. Sarah Laframboise, executive director of Evidence for Democracy, noted a pervasive lack of human oversight in administrative policy drafting. Political leaders in New Brunswick and beyond face mounting pressure to establish robust disclosure frameworks before automated drafting standardizes legislative discourse.