Canadian Lawmaker Reads AI Prompt Instructions Aloud in Floor Speech as Viral Video Sparks Debate

Bill Oliver was speaking last month in the New Brunswick legislature about government accountability. Citizens, he told colleagues, often develop expectations that exceed the powers actually granted to advocacy offices. Then he continued with a line no human speechwriter would produce: "here's a more natural, flowing version of that section that reads like a legislative speech rather than a series of short points." 1

That line is meta-commentary. It is the kind of instruction a chatbot generates when you ask it to revise a passage, explaining how it restructured your text before presenting the result. The expectation is that you delete that explanation. Oliver, a Progressive Conservative Party member of New Brunswick's legislative assembly, read it into the legislative record apparently without recognizing what it was. 1

The speech drew no attention at the time. Video of the remarks started circulating on Reddit and Threads earlier this week, and within days the CBC and the Toronto Star had picked up the story. The Star characterized the episode as reflecting a class divide between people comfortable delegating their work to AI and citizens who object to it. 1

Oliver is the latest entry in a growing catalogue of professionals whose AI use has surfaced publicly and awkwardly. Lawyers have submitted court filings citing cases that do not exist, fabricated by language models that invent precedents with confident specificity. Academics have published papers containing AI-generated text with fabricated references. Journalists have filed work whose machine origins became impossible to deny once readers noticed the tells. 1

Each incident gets treated as an embarrassment, a one-off lapse by someone who should have been more careful. But workers who use AI tools tend to hide that fact from colleagues, according to a Duke University study, because colleagues perceive AI-assisted work as evidence the person is "lazy" or "replaceable." 1

Oliver used AI to help draft a legislative address, which is neither unusual nor necessarily wrong. His mistake was failing to strip the model's scaffolding before he stood up. People skip that cleaning step for the same reason they hide the tool entirely: in most professional environments, admitting you used AI carries a cost unrelated to the quality of the work. A well-written speech produced with AI assistance still invites suspicion. A poorly written speech produced without it does not. The incentive structure punishes honesty about the process.

The cost exists because the norms have not been built. There is no established expectation that a legislator disclose AI-assisted speechwriting. No rule requiring a lawyer to flag that a filing was drafted with a language model. No standard for when a published paper should mention AI assistance in its methods section. In the absence of norms, every voluntary disclosure functions as an admission of inadequacy rather than a transparency measure. Professionals use the tools quietly, a few get caught, and the public reacts to each exposure as though it were an aberration rather than the predictable byproduct of a workforce using tools it is expected to pretend it does not need.

The people writing laws, filing lawsuits, and publishing research are integrating AI into their daily work. Most will not say so. The ones who get caught are not the most reckless. They are the least careful about concealing it. Oliver read his AI's instructions aloud because he did not recognize them. Other professionals across legislatures, law firms, and university departments are doing the same thing more quietly. The only difference is that nobody has posted their video.

Every accidental exposure treated as a scandal reinforces the impulse to hide. The hiding guarantees more careless mistakes, which guarantees more scandals. The loop tightens because no institution has stepped forward to say what everyone watching these stories already suspects: that AI use among professionals is not exceptional. It is routine, it is growing, and the shame attached to it is doing more damage than the tools themselves ever could.