What happened
On 14 April 2026, The New York Times ran a story about Canadian politics. It quoted Conservative leader Pierre Poilievre, from a March speech, saying politicians who switch parties should “resign their seats tonight and run in a by-election tomorrow.”
He hadn’t said that. The words came from an AI tool the reporter used, which had summed up his views and presented the summary as a quote.
A few weeks later, the paper added an editors’ note admitting the mistake and swapping in a real quote from an April speech.
Why it went wrong
AI tools that summarise speeches can write the summary in the speaker’s voice, inside quote marks. It looks exactly like a real quote. The editors’ note was clear: “The reporter should have checked the accuracy of what the AI tool returned.”
Impact
- One of the world’s best-known newspapers put words in a politician’s mouth.
- The fix came more than two weeks after the story ran.
The lesson
- A quote must come from the source. Check the video, the transcript or the speech itself, never a summary.
- Quote marks mean “these exact words.” AI tools don’t always respect that.
- If it can happen at the Times, it can happen anywhere.