Canadian politician accidentally read a ChatGPT response aloud during an official speech

A bizarre incident in Canada quickly spread across social media, reigniting debate over how heavily politicians are beginning to rely on artificial intelligence. During a speech, Bill Oliver, a member of New Brunswick’s Legislative Assembly, unexpectedly read aloud a phrase that was clearly not intended for the public: “Here is a smoother version of this section, written as a full legislative speech rather than a series of short talking points.” He then continued reading the text without pausing, as if nothing unusual had happened.

Users quickly guessed that the politician had probably accidentally read out an internal note left in the document after working with ChatGPT or another AI assistant. Most remarkably, neither Oliver nor those around him in the chamber immediately noticed the mistake. It was only after a recording of the session was published that the moment went viral, with the clip attracting thousands of views and triggering a wave of jokes and criticism online.

The incident became another example of how generative AI is making its way into politics, law and public administration. Experts note that using neural networks to prepare drafts has already become common practice, but such tools require mandatory human review. Otherwise, internal comments, instructions or even fabricated facts can end up in official documents and public speeches.

On social media, users joked that politicians were beginning to “delegate their thinking” to neural networks, while some commentators compared the situation with recent cases in which lawyers submitted documents containing nonexistent legal precedents generated by AI. The story of Bill Oliver once again showed that even the most convincing speech requires not only good writing, but also a careful final check by a human.