Canadian Politician Reads AI Prompt Aloud in Legislature

A Canadian lawmaker just became an unintentional case study in why AI-assisted writing needs a careful final read-through before it reaches a live audience. Bill Oliver, a Progressive Conservative member of the Legislative Assembly of New Brunswick, read an apparent AI chatbot prompt aloud during a floor speech last month, and the moment has since gone viral across Canadian social media, according to Al Jazeera's coverage of the incident.

Oliver was addressing the potential pitfalls of creating advocacy offices when the mix-up happened. "One of the dangers associated with creating advocacy offices is that citizens often develop expectations that exceed the powers actually granted to those offices," he said, according to Breitbart's reporting on the transcript. He then continued directly into what was clearly meant to be an internal editing note, not part of the speech itself: "Here's a more natural, flowing version of that section that reads like a legislative speech rather than a series of short points."

The phrase Oliver read aloud is instantly recognizable to anyone who has used an AI writing assistant. It's the kind of transitional language a chatbot commonly generates when offering an alternative version of a passage in response to a prompt, exactly the sort of instruction meant to sit above the actual content, not be spoken as part of it. Footage of the moment was posted to the New Brunswick subreddit and quickly drew significant attention, with commenters comparing the error to professional standards expected in other fields and some even calling for Oliver's resignation, according to 1News's reporting on the online reaction.

Oliver's gaffe isn't an isolated embarrassment. It follows a string of similar incidents across professional fields where AI-generated content made it into final output without adequate review. Two lawyers were sanctioned by a U.S. appeals court last month for filing documents riddled with AI hallucinations, and four additional lawyers from both sides of a Mississippi lawsuit faced separate sanctions for the same underlying problem, with two barred from appearing in federal court for the next two years, according to Futurism's reporting on the broader pattern. These incidents connect directly to concerns we've covered in our guide on AI hallucinations and how they happen, where confidently wrong or improperly reviewed AI output causes real professional consequences.

In my four years in sales at a research and advisory firm, I heard directly from CMOs and CEOs about what they wanted from AI, and the fear of exactly this kind of embarrassing public mistake came up constantly, especially from leaders responsible for public-facing communications, board presentations, or client-facing documents. Oliver's viral moment is a genuinely useful, low-stakes cautionary example for any business using AI to draft speeches, presentations, or official communications.

The practical lesson is simple but easy to skip under deadline pressure. Any AI-assisted content headed for a public or client-facing audience needs a dedicated final review specifically checking for leftover instructional language, editing notes, or prompt artifacts that were never meant to be seen, not just a general proofread for typos and tone.

New Brunswick legislator Bill Oliver accidentally read an AI chatbot's editing instructions aloud during a legislative speech, mistaking the AI's transitional phrasing for part of his actual remarks. The moment went viral on Canadian social media after footage was posted to Reddit, drawing both mockery and calls for accountability. The incident joins a growing list of professional AI mishaps, including recent court sanctions against lawyers for filing AI-hallucinated legal documents.