Canada’s new AI for All national strategy, according to Prime Minister Mark Carney, ensures that Artificial Intelligence will serve all Canadians, “not the other way around.” The strategy, introduced in June 2026, outlines ways AI models can build Canadians' trust, create economic opportunity, and protect sovereignty from the United States. Loosely tied to multiculturalism, the strategy claims AI must strengthen French-language use by “capturing and projecting its idioms, expressions, and cultural contexts.”
Although AI use carries legal ramifications across Canada, its role in French-language protections in and outside Quebec is especially noteworthy. According to the Canadian Association of Professional Employees (CAPE), over 90% of translations in the federal government are from English to French. As AI models increasingly perform these translations for individuals, businesses, and legal officials, French speakers throughout Quebec and Canada bear the brunt of AI-translated work.
This creates serious ramifications relating to Quebec’s Bill 96, which made French the sole language of communication between government officials and Quebecers. Bill 96’s French-language mandates become increasingly convoluted when AI translation is involved. According to The Montreal Gazette's Jack Jedwab, shifting to instant translation alters the bill from changing the language used in the public sphere to a superficial use of publication language that affects the appearance of government bodies rather than individual behaviour. Although this can increase access to government websites for many non-French speakers, it ultimately undermines many of the bill's purposes and Quebec’s language protections more broadly. Jedwab claims that AI will expose the limits of many of Quebec’s language laws, as politicians face increasing pressure to explain whether language protections are purely symbolic or producing measurable changes in French language use.
Jedwab’s argument is evident in the advertising of many AI chatbots. Smartcat advertises its translation services specifically for Canadian government translation from English to French. Meanwhile, AI translation platform Lilt advertises specifically around Bill 96, offering compliance webpage and communications services to business owners. AI has become integral to French translation and compliance, with the Quebec government supporting many AI measures. The federal AI for All framework states that any AI systems deployed for government use must function equally well in both languages, and Quebec has invested billions of dollars in AI projects.
However, the AI models most commonly used in business and government communications are still largely English-dominated. 80-90% of training data for ChatGPT, Claude, and Gemini is in English. Although a small portion of the training data is in European French, the models generally perform 15 to 20% worse in Quebec French than in European French. Because the province’s sovereignty is largely based on regionally specific French, the sovereignty benefits AI offers to Canada more broadly may not apply in Quebec. Excessive use of AI models in business, government, and personal communications risks blending the Quebec dialect with European French, or even Canadian English.
Additionally, AI use can pose further trouble for legal and judicial translation. The Quebec legal guidelines on the use of artificial intelligence for judges do not specifically ban AI use in the courtroom. Although the models are banned from use in judicial interpretation and rulings, most of the guidelines are recommendations, especially for lawyers. AI models are not expressly banned from citing witness reports or jurisprudence, which raises concern given the model’s poor Quebec language performance. Furthermore, despite outstanding performance on the American bar exam, ChatGPT earned a final score of 12% on the Quebec bar exam. AI models are being normalized in Quebec’s government and legal communication, yet they show low proficiency in both Quebec French and the province's legal standards.
Quebec has furthermore invested millions in AI for medical systems and data processing. This has drawn heavy criticism, as most Quebec hospitals still rely on handwritten notes and fax machines, lagging far behind the rest of Canada. Because these archaic systems limit Quebec medical data compared to other provinces, these models will likely again be trained on non-local data. As a result, they may perpetuate inaccuracies within the Quebec medical system, harming Quebecers seeking care.
As AI models are increasingly advertised for their accuracy, flexibility, and translation abilities, it remains pertinent for Quebec to assess how AI relevance in the province may differ from the broader Canadian context. The federal and provincial governments have embraced AI in government, corporate, legal, medical, and interpersonal contexts. However, because the models are primarily trained on data from outside Quebec, many of their legal, medical, and linguistic interpretations are likely to perpetuate inaccuracy and harm. Quebec’s Bill 96 established a broad standard for French communication and protections; however, limited AI regulation in subsequent acts threatens that very promise. Quebec Bills such as C-34, addressing digital safety, and C-36, addressing data privacy, are viewed in broader Canada as a targeted approach to AI; however, they provide few of the linguistic protections mandated by Bill 96. To fulfill the promises outlined in AI for All, promoting sovereignty, trust, opportunity, and strengthening of the French language, Quebec must further its AI regulations beyond the all-encompassing standard set by Bill 96. Artificial Intelligence must not be treated as a solution to linguistic divides, as it is not yet capable of replicating the distinctions of Quebec’s language, laws, and culture.
Edited by Emily Nooranbakht