Canada's National AI Strategy Lands Within Days as Ottawa and TELUS Advance Sovereign AI Infrastructure in British Columbia
On Wednesday May 27, Prime Minister Mark Carney told reporters on Parliament Hill that the federal government's long awaited national AI strategy will be released within days. The spring economic statement set out six pillars: new privacy and online safety laws, sovereign compute infrastructure, support for Canadian AI companies, coordination with international allies, AI training and education, and pro worker industrial AI. The strategy is paired with concrete infrastructure. On Monday May 25, AI Minister Evan Solomon announced that TELUS will build a new AI data centre in British Columbia under the federal Enabling Large Scale Sovereign AI Data Centres initiative, drawing on an initial 85 megawatts of clean power and housing more than 60,000 graphics processing units once operational.
The throughline for enterprise leadership is the distinction between where data sits and who controls the call path. Sovereign compute on Canadian soil addresses residency, but governance, model provenance, and operational control of the inference path determine whether an AI deployment is genuinely defensible. The strategy signals a sustained federal posture pairing capital for sovereign infrastructure with privacy and online safety law reform, which together reshape both the supply of Canadian AI capacity and the rules under which it operates.
- Carney said on May 27 the national AI strategy will be released within days
- Six pillars include sovereign compute infrastructure and modernized privacy and online safety laws
- TELUS to build a British Columbia AI data centre under the federal sovereign AI data centre initiative, announced May 25
- Initial 85 megawatts of clean power, more than 60,000 GPUs once operational
Enterprise Impact: CIO, CTO, and data leadership should map current and planned AI workloads against Canadian sovereign deployment lanes, distinguishing data residency from operational control of the inference path. Procurement should ask AI vendors for explicit Canadian residency, model provenance, and data egress terms, and align AI governance to ISO/IEC 42001:2023. Boards should expect sovereign compute provenance, Canadian residency, and the privacy reform carried inside the strategy to feature in federal procurement and customer due diligence through the rest of 2026.
Source: Government of CanadaBetaKit Profiles Cohere as Canada's AI Champion, Framing Enterprise Deployment Across Telco, Financial Services, Energy, and the Public Sector
BetaKit published a profile on May 26 framing Toronto based Cohere as Canada's chosen AI champion and a load bearing part of the national sovereign AI position. Unlike consumer chatbot makers, Cohere builds enterprise grade language models designed to be deployed privately on a customer's own infrastructure, with adoption concentrated in regulated and infrastructure heavy sectors including telecommunications, financial services, energy, and the public sector. The profile highlights Cohere's North enterprise platform, its Bell partnership, and the federal sovereign AI compute support behind the company.
- BetaKit profiled Cohere on May 26 as Canada's AI champion within the national sovereign AI position
- Cohere builds enterprise models for private deployment on customer infrastructure, not consumer chatbots
- Adoption focus spans telecommunications, financial services, energy, and the public sector
- Profile highlights the North enterprise platform, the Bell partnership, and federal sovereign compute support
Enterprise Impact: The signal for Canadian enterprises is that practical AI adoption is moving toward privately deployed, sector specific models where residency, weight ownership, and integration into existing operations matter more than general purpose chat. Leadership should evaluate enterprise AI options on deployment locality, data control, and integration depth rather than demo quality. Procurement should request reference deployments in the relevant sector and explicit terms on data handling and model provenance, with ISO/IEC 42001 alignment anchoring the governance wrapper around any production deployment.
Source: BetaKitCalian and Cohere Partner to Bring Sovereign AI Into Defence Operations, Applying the North Platform to Decision Support, Training, and Readiness
On May 28, Calian Group and Cohere announced a collaboration to evaluate and integrate sovereign, Canadian developed enterprise AI in defence environments. The partnership applies Cohere's secure agentic platform North within controlled defence settings to support faster operational and training insights for the Canadian military, allied forces, and small to mid sized enterprises through Calian Ventures, without exposing sensitive data to public networks. Calian also debuted its ATHORA interoperability platform alongside the Cohere partnership at the CANSEC defence trade show.
- Calian and Cohere announced a sovereign AI defence collaboration on May 28
- Cohere's North platform to be applied to decision support, training, and operational readiness in controlled environments
- Targets the Canadian military, allied forces, and SMEs via Calian Ventures, keeping data off public networks
- Calian debuted its ATHORA interoperability platform with the Cohere partnership at CANSEC
Enterprise Impact: The partnership is a concrete example of AI moving from pilots into regulated, high assurance operational workflows where data isolation and provenance are mandatory. Organizations in regulated or security sensitive sectors should treat it as a template: secure deployment boundaries, controlled data paths, and clear human oversight are prerequisites, not enhancements. Leadership evaluating AI for sensitive operations should require documented isolation, audit trails, and ISO/IEC 42001 aligned governance before scaling beyond a pilot.
Source: SpaceQFedDev Ontario Commits Nearly $16.5M to 13 Greater Toronto Area AI Companies, Backing Adoption in Regulated and Operational Workflows
The Government of Canada, through FedDev Ontario, announced on May 25 nearly $16.5 million for 13 Greater Toronto Area companies and organizations advancing Canadian made AI technologies and adoption. The recipients map cleanly onto regulated and operational use cases: Private AI, operating as Limina, for sensitive data infrastructure used by regulated enterprises in healthcare, financial services, and insurance; DMD Building Systems for integrating robotics and AI into engineering workflows, including extracting structured data from non standard formats; and VisFuture for connecting operational data across SME functions, alongside recipients including the Vector Institute, Future Fertility, and ProteinQure.
- FedDev Ontario announced nearly $16.5 million on May 25 for 13 GTA AI companies and organizations
- Funding targets AI adoption and faster commercialization across regulated and operational sectors
- Private AI, operating as Limina, backed for sensitive data infrastructure in healthcare, financial services, and insurance
- DMD Building Systems and VisFuture funded for AI in engineering workflows and SME operational data access
Enterprise Impact: The recipient mix is a useful map of where Canadian AI value is concentrating: regulated data readiness, document and data extraction, and operational data integration rather than generic content generation. Enterprise leadership should read it as confirmation that AI adoption returns most where it connects to existing operational data and governed workflows. Organizations in regulated sectors should prioritize data readiness, governance, and integration capacity before model selection, with ISO/IEC 42001 aligned governance and documented data handling accompanying any sensitive data deployment.
Source: Government of CanadaCohere and Mila Partner on Quebec French Language and Cultural Context in AI, Opening a Montreal Office to Advance Multilingual Evaluation
On May 27, Cohere and Mila announced an academic research collaboration to advance multilingual and multicultural AI evaluation, beginning with French language cultural context in Quebec, and Cohere announced a Montreal office tied to the work. The collaboration focuses on how frontier models can move beyond standardized outputs to reflect the cultural, social, and institutional context that shapes how Quebec French is used in practice, and on building the evaluation methods organizations need to assess AI performance in high context, multilingual environments.
- Cohere and Mila announced a research collaboration on May 27 on multilingual and multicultural AI evaluation
- Initial focus on Quebec French language and cultural context
- Cohere is opening a Montreal office tied to the collaboration
- Goal is stronger evaluation methods for AI performance in high context, multilingual settings
Enterprise Impact: For Canadian organizations operating in French and English, the collaboration underscores that AI adoption is not only a technical exercise; language accuracy, cultural context, and evaluation rigour determine whether a deployment is trustworthy in regulated or public facing settings. Leadership deploying AI in bilingual or Quebec contexts should require documented evaluation across languages and cultural context, not only aggregate benchmark scores. Procurement should ask vendors how models are evaluated for Quebec French and other in market languages before production use.
Source: MilaAnthropic Expands Its Enterprise Footprint Internationally, Opening a Milan Office and Naming a Korea Representative Director Ahead of a Seoul Office
Anthropic continued an international enterprise build out in late May. On May 27 the company opened a Milan office, its sixth in Europe, to support Italian enterprise, research, and developer customers, with early engagements spanning finance, life sciences, energy, and automotive. On May 26 it appointed KiYoung Choi as Representative Director of Korea ahead of opening a Seoul office, with the Korea team focused on enterprise and startup partnerships, government and research engagement, and the local developer community.
- Anthropic opened a Milan office on May 27, its sixth in Europe, targeting Italian enterprise and developers
- Early Italian engagements span finance, life sciences, energy, and automotive sectors
- Anthropic appointed KiYoung Choi as Representative Director of Korea on May 26 ahead of a Seoul office
- The expansion signals continued enterprise and international growth among frontier AI providers
Enterprise Impact: The pattern across frontier AI providers is direct enterprise and public sector engagement in local markets rather than purely remote API access. Enterprise leadership should expect more localized support, partnership, and procurement pathways from major AI vendors, and should weigh local presence, data residency, and support commitments alongside model capability when selecting providers. Procurement should ask about in region support, data handling, and contractual terms for the jurisdictions where workloads actually run.
Source: AnthropicAnthropic Releases Its Latest Frontier Model, Opus 4.8, Adding Dynamic Agentic Workflows and a Cheaper High Speed Mode for Enterprise Use
On May 28, Anthropic released Opus 4.8, an upgrade to its most capable model line that builds on the prior version with gains on agentic benchmarks and reliability at the same price point. The release adds a dynamic workflows capability for large scale, multi step tasks in the company's coding product, and a high speed mode that runs at roughly 2.5 times the speed at about a third of the previous cost. Listed pricing is $5 per million input tokens and $25 per million output tokens, with further savings available through prompt caching and batch processing.
- Anthropic released the Opus 4.8 frontier model on May 28 at the same price as the prior version
- Adds dynamic workflows for large scale, multi step tasks and a cheaper high speed mode
- Reported gains on agentic task benchmarks and reliability over the prior model
- Listed pricing of $5 per million input tokens and $25 per million output tokens, with caching and batch savings
Enterprise Impact: Frontier model capability and cost continue to move on a months long cadence, which means enterprise AI architecture should assume model swappability rather than lock in. Leadership should design AI systems with an abstraction layer over model providers, evaluate new releases on agentic reliability and total cost rather than headline benchmarks, and revisit build versus buy decisions as inference costs fall. Procurement should keep model contracts flexible enough to adopt newer releases without re platforming.
Source: TechCrunch