Canadian Politician Accidentally Reads AI Chatbot's Own Instructions Aloud in New Brunswick Legislature

A Canadian provincial politician has become the latest example of an unedited AI-generated passage slipping into a formal public setting, after video surfaced showing him reading a chatbot's own instructional text aloud during a speech to his province's legislature.

A mid-speech digression that went unnoticed

Bill Oliver, a Progressive Conservative member of the Legislative Assembly of New Brunswick who represents the Kings Centre district, was addressing a matter involving the Office of the New Brunswick Advocate during a floor speech last month. Partway through his remarks, Oliver appeared to read directly from what looked like an AI chatbot's formatting note rather than his own prepared remarks. "Here's a more natural, flowing version of that section that reads like a legislative speech," Oliver said, a line widely recognized as the kind of meta-instruction large language models often include when presenting a rewritten passage back to a user, rather than content meant to be spoken aloud as part of the finished text.

The moment came shortly after Oliver had delivered a substantive point about the risks associated with creating new government advocacy offices, noting that citizens often develop expectations that exceed the powers actually granted to those offices. That portion of his remarks read as a standard piece of legislative rhetoric. It was the section immediately following it — the apparent AI formatting note — that drew attention once footage began circulating widely online.

From unnoticed gaffe to viral clip

Although the speech was delivered last month, the moment went largely unnoticed at the time. Footage of the exchange was posted to the New Brunswick subreddit this week and quickly began circulating on Reddit and Threads, drawing criticism from users who said the mix-up reflected poorly on the quality of preparation behind the speech. The clip was subsequently picked up by the Canadian Broadcasting Corporation and the Toronto Star, which covered the moment as part of a broader conversation about how political and professional figures are increasingly relying on AI tools to draft public remarks, sometimes without adequately reviewing the output before delivering it.

Commenters reacting to the footage compared the standard expected of Oliver to the level of scrutiny typically applied to professionals in fields like law, where similar AI-related mistakes have drawn public criticism in the past. Some online commenters went further, calling for Oliver to resign over the incident, though no formal calls for his removal have been reported through official channels within the New Brunswick legislature.

Part of a broader pattern of AI slip-ups

Oliver's moment adds to a growing list of instances in which unedited AI-generated text has slipped into professional or public-facing work, spanning fields including law, journalism, academia and now provincial politics. Observers who track these incidents note that the phrases most likely to expose AI involvement tend to be meta-commentary the tool generates about its own output, such as offering "a revised version" or "a more conversational tone," language intended for the person prompting the AI rather than for a public audience.

Some researchers who study workplace AI use have pointed to broader cultural dynamics behind these kinds of incidents. Prior research on professional AI use has found that workers frequently avoid disclosing their use of AI tools out of concern that colleagues might view the practice as lazy or as a sign that their work could be easily replaced, a dynamic some analysts say may contribute to insufficient review of AI-assisted material before it is finalized or delivered publicly.

New Zealand grapples with similar questions, without a similar mishap

The incident in New Brunswick has drawn comparisons to an unrelated but related conversation playing out in New Zealand's Parliament, where officials have been considering how AI tools should factor into official submissions, though not because of any similar public misstep. David Wilson, Clerk of the New Zealand House of Representatives, said there was nothing inherently wrong with people using AI tools to help refine their written submissions to Parliament, so long as the views ultimately expressed remained genuinely their own. According to Wilson, the central concern for lawmakers and officials isn't the technology itself, but whether a given submission still reflects a person's own thinking rather than content generated wholesale by an AI system.

No indication of any rule violation

Coverage of the New Brunswick incident has generally framed the episode as a question of accountability and review quality rather than a breach of any specific legislative rule. No formal violation of assembly procedure has been reported in connection with Oliver's remarks, and the broader criticism directed at him has centered on the appearance of insufficient preparation rather than any claim that using AI tools to draft political speeches is itself against the rules.

A reminder about reviewing AI output

The episode has been widely cited as a cautionary example for professionals across fields who use AI tools to help draft written or spoken material. Commentators covering the incident have generally offered the same basic takeaway: AI-generated text should always be reviewed in full before being published or spoken publicly, since large language models frequently embed their own formatting notes, alternative phrasing options or other meta-commentary directly within their output, material that is meant to guide the person using the tool rather than to be presented as finished content.

As of this week, Oliver had not issued a public statement addressing the footage or explaining how the apparent AI-generated line made its way into his prepared remarks. The Legislative Assembly of New Brunswick has likewise not indicated whether the incident will prompt any formal review of how AI tools are used in preparing official speeches or submissions. For now, the episode continues to circulate as one of the more visible recent examples of an unedited AI artifact reaching a public, official setting — a reminder, as several outlets covering the story have noted, that even brief lapses in review can turn an ordinary AI-assisted speech into an unintended viral moment.