L’égalité réelle des langues officielles : un angle mort de L’IA pour tous

Artificial intelligence (AI) is profoundly transforming social relations, research activities, and the economy. Everyone must take part in this technological shift, and in this regard, postsecondary institutions in minority francophone settings play a strategic role in developing an inclusive, bilingual, and discoverable digital ecosystem.

Yet in public discourse on AI in Canada, little is said about the scope of substantive equality between the two official languages, a principle at the heart of the Official Languages Act (OLA). Moreover, the AI for All strategy only aims for equivalent performance of AI tools in both official languages, which we believe falls short of respecting the spirit of the OLA.

In itself, the OLA encourages the application of a linguistic lens in designing positive measures aimed at equitable AI adoption by official language minority communities (OLMCs). These measures must take the realities of OLMCs into account. Doing so would allow the strategy to capture and value francophone cultural contexts, maximize AI adoption, and ensure that its benefits are concrete for the entire Canadian population.

Why must a linguistic lens be applied in implementing AI for All? We highlight three reasons.

1. Equitable participation and consultation of OLMCs

French is not simply a language for translation, and AI adoption in Canada should not relegate either the French language or francophone perspectives to a secondary role. At present, English dominates technology development. So when the strategy refers to equivalent performance of tools in both languages, what is really meant is equivalent French-language performance of tools developed in English.

However, substantive equality between the two official languages calls instead for the equivalent development of tools in both French and English. The design of AI tools and training in their use should be carried out in a way that fosters equitable participation and consultation of OLMCs, which is a cornerstone of the OLA. OLMCs must be able to present their local realities regarding infrastructure, data, and training needs, to ensure adequate and equivalent AI adoption throughout the educational continuum and in the francophone labour market.

Fostering equitable participation and consultation would also help address linguistic biases in data, corpora, algorithms, and their use. Together, these factors tend to exacerbate socioeconomic inequalities. Translating outputs is not enough to resolve epistemic issues — it merely reproduces existing inequalities. Awareness and training on linguistic bias are crucial to designing reliable, high-quality models that serve OLMCs.

To achieve this, any measure arising from the federal AI strategy should incorporate the key principles of the OLA from design through to evaluation, to ensure the development of models suited to OLMC realities. In this respect, postsecondary institutions in minority francophone settings are valuable partners in deploying the federal AI strategy.

2. Support for research and skills development in French

The recent report from the External Advisory Group on the Creation and Dissemination of Scientific Information in French highlights the normative, financial, and organizational challenges associated with French-language research in Canada. These challenges appear across all fields of knowledge, including AI development.

To ensure the active participation of OLMCs in the strategy, the federal government must aim for a more equitable distribution of research funding — including funding for research infrastructure — in order to build the capacity of colleges and universities in minority francophone settings. These institutions work directly with civil society, institutional partners, and businesses to address emerging needs and promote broader access to AI. Basic and applied research in French would also support the development of the professional and technical vocabulary needed to build reliable AI tools of equivalent quality. But this requires infrastructure that allows for the collection, management, and use of data on OLMCs — data that remains scarce and is not systematically gathered.

In addition, the federal AI strategy relies on access to training for the entire population. Building digital literacy is essential in a rapidly changing environment. Yet digital literacy is acquired alongside language skills. To make AI accessible and encourage its adoption, training adapted to linguistic specificities — including a critical approach and awareness of linguistic bias — is essential. Simply translating training materials developed in English is not enough to embody substantive equality between the official languages in this project; rather, an approach designed by and for OLMCs is needed.

3. Canada’s leadership within the international Francophonie

For Canada to become a global leader in AI, it is well-positioned to take a leadership role within the international Francophonie. The evolution of AI illustrates the importance of global collaboration in developing AI strategies, standards, and models that are inclusive and adapted to local realities. In this regard, Canada’s linguistic framework allows it to stand out through its expertise in French-language AI, which could be shared and commercialized internationally.

By stepping up its efforts in scientific diplomacy, Canada can play a crucial role in positioning itself — all the more so if it hosts the 21st Francophonie Summit in 2028. Greater involvement would showcase Canadian expertise and the key role of postsecondary institutions and civil society in AI research. Yet the federal AI strategy overlooks such opportunities.

In short, to embody the spirit of the OLA, the federal government must do more than what is currently planned under AI for All. Substantive equality between the two official languages must permeate every stage of its implementation. This is one clear path to ensuring that this strategy is truly for everyone.