Ontario lawmakers have unanimously endorsed greater transparency for AI-generated advertising, but a new investigation has exposed a gap in the province’s approach to artificial intelligence: the political speeches delivered by those same lawmakers face no equivalent disclosure requirement under existing provincial AI rules.
On October 28, the legislature approved Progressive Conservative MPP Silvia Gualtieri’s motion by 116 votes to zero, urging the government to pursue measures identifying artificial intelligence used in commercial advertisements and product representations.
The motion was advisory rather than binding legislation.
Sponsored · Selkirk Copper Mines
Ontario already imposes AI transparency requirements elsewhere in government. Its Responsible Use of Artificial Intelligence Directive, effective December 1, 2024, covers ministries and provincial agencies using AI in policy development, programs, and public services.
Those institutions must disclose AI use in certain public-facing interactions and decisions affecting individuals. The directive does not explicitly extend those requirements to speeches delivered by elected legislators.
The disclosure gap gained attention after a recent investigation by Sabrina Nanji of the Queen’s Park Observer. Nanji analyzed nearly 400,000 words from seven legislative sitting days between October 27 and November 5, using the Pangram AI-detection platform.
The software classified 14.8% of the words as AI-generated or likely AI-generated.
Sponsored · Cambria Gold Mines
The analysis reported substantial differences between political parties:
- Progressive Conservatives: 21%
- Liberals: 15%
- New Democrats: 8%
- Greens: 0%
Nanji also reported that 51 MPPs had at least one statement receiving an AI-detection score of 5% or higher.
A comparison against seven sitting days from 2022 produced no AI flags.
The investigation also examined Gualtieri’s own remarks advocating greater transparency in advertising. According to Nanji, Pangram classified every word of the MPP’s opening and closing remarks as AI-generated.
The official legislative transcript confirms that Gualtieri delivered the remarks, but does not establish whether AI was involved in writing them. During the October 27 debate, Gualtieri argued that consumers should understand when artificial intelligence influences commercial content.
“When you are shown something in order to sell you something, you deserve to know what you are looking at,” she told lawmakers.
The same principle raises questions about whether voters should receive comparable information when elected representatives deliver prepared arguments.
The investigation included denials from legislators’ offices. A spokesperson for Progressive Conservative MPP Will Bouma said neither he nor his team had used AI to generate speeches or statements.
The NDP said it was developing internal guidelines for responsible AI use in consultation with its staff union.
It is worth noting AI-detection software cannot conclusively identify who wrote a speech or distinguish every instance of automated drafting from human writing. A 2026 peer-reviewed evaluation found Pangram performed strongly on controlled test documents, but cautioned that detection results should not serve as the sole evidence in high-stakes decisions.
The findings therefore establish neither misconduct nor a breach of Ontario’s existing AI directive.
