New York Times retracts AI-generated quote misattributed to Canadian politician

The New York Times published a correction acknowledging that a statement attributed to Conservative Party leader Pierre Poilievre in a Canadian election story was not an actual quotation but rather text generated by an AI system. The tool had produced what appeared to be a coherent summary of Poilievre's political positions, and the reporter treated it as a direct quote without verification.

The editors' note explicitly faulted the reporting process, stating the journalist should have validated the AI tool's output independently. The fabricated remark—describing politicians who switched allegiances as 'turncoats'—did not appear in Poilievre's actual April speech that the article was supposed to cover.

The incident illustrates a systemic risk: language models generate confident-sounding text that can pass casual inspection, especially when trained on domain-specific information. In journalism, where direct attribution carries legal and ethical weight, deploying these tools without mandatory fact-checking creates a vector for unintentional misinformation at scale.