---

Principal Authors

---

Contributors

• Catherine Amburgey
  Zaynab Choudhry
  Thomas Goldsmith, Orbit Policy
  Marium Hamid
  Jake Hirsch-Allen
  Angus Lockhart
  

---

Tags

• Digital Economy
• Digital Government
• Digital Skills
• Future of Work
• Inclusive Growth
• K-12 Education
• Non-Profit Sector
• Scale-ups and Start-ups
• Skills Training

---

Share

Responsible AI Adoption Built on Trust and Effective Governance

---

AI deployment requires high upfront costs, from investments in tools and applications, to computing capacity, and worker training and reskilling. Alongside firm-level uncertainty about AI’s benefits for business processes and productivity, achieving broad adoption depends on addressing common organizational barriers and capacity challenges. It also requires efforts to build public and workforce trust that AI will be developed and deployed responsibly. 

In this section, we describe the state of AI adoption in the Canadian economy, and outline recommendations for the strategy to advance adoption. We then outline why public trust must be central to Canada’s AI agenda, and propose mechanisms for trust-building through AI governance.

---

THEME 1

AI Adoption Across Industry and Governments

Private sector adoption

Despite the hype about AI, Canada’s private sector has been slow in adopting the technology. By 2021, our research, Automation Nation, found that just under 4 percent of Canadian firms had incorporated AI into at least one business function. By 2023, after the launch of ChatGPT and the mainstreaming of generative AI, this rate had only increased to nearly 7 percent. Importantly, adoption is highly concentrated in large companies (25.9 percent compared to 5.9 percent in small firms in 2023) and uneven across industries: information and cultural industries have the highest rate of adoption (25.6 percent), followed by professional and technical services (14.7 percent), and finance and insurance (10 percent). 

The productivity impacts of firm adoption have also been muted to date. Our research, Waiting for Takeoff, analyzed firm-level AI adoption in the 2020 to 2022 period (preceding the mainstreaming of generative AI) and found that businesses that had adopted AI did not experience short term productivity gains. More recent work focusing on the productivity impact of generative AI reached the same conclusion.

Our research identifies two key adoption barriers. First, the vast majority of AI non-adopter firms cite not having any use for the technology as the primary factor for non-adoption. Second, companies that offer ICT training (to both technical and non-technical staff) are much more likely to adopt AI. We thus believe that knowledge gaps are the key barrier to adoption. AI is not adopted because key decision makers (and line employees) have insufficient levels of AI literacy to navigate AI integration processes, which can then hinder uptake within a company’s broader digital culture, as shown in our report Picking Up Speed.

Finally, while much discussion on AI adoption focuses on for-profit companies, the absence of focus on Canada’s nonprofit sector needs to be addressed in the new strategy. Canada’s nonprofit sector, representing over 170,000 organizations, is critical to the country’s economy and social fabric. In a sector that is often under-resourced and faces digital skills and technology gaps, AI offers significant potential, as outlined in our reports Canada’s Nonprofit Tech Workforce and The Demand for Digital Skills in Canadaʼs Nonprofit Sector. Pilot initiatives to offer bespoke training, frameworks and support, like the Responsible AI Adoption for Social Impact (RAISE) delivered by the Dais with partners and support from DIGITAL, offer great scale-up potential. 

Public sector adoption

The government of Canada has prioritized public sector AI adoption to improve operations and achieve spending targets through efficiencies. The G7 statement on AI specifically mentioned the launch of the Rapid Solutions Lab to “develop innovative and scalable solutions to the barriers we face in adopting AI in the public sector.”

Conversations on government AI adoption in the public sector cannot occur outside of broader conversations on its digital maturity. Once a leader in the United Nations’ E-Government rankings, Canada’s ranking tumbled to 47th place by 2024. Our study of digital transformation in the federal government, Byte-Sized Progress, found persistent culture and talent barriers to incorporating new technologies. 

Therefore, we believe the new AI strategy must build upon existing digital government efforts and directives on AI use in government, including the AI Strategy for the Federal Public Service, a recently-updated Directive on Automated Decision-Making for government use, and Digital Competencies for Public Servants, and focus on effectively addressing AI within these directives.

Our recent study Adoption Ready? found a much higher concentration of federal public sector workers in highly AI-exposed occupations, composed of tasks that can be automated, than across the general Canadian workforce (59% compared to 29%). We don’t believe that 59% of federal jobs can or should be automated, but it highlights the potential scale of impact that decisions around AI adoption in government may have. 

AI adoption may present opportunities, but also the potential for disruption for public servants and government services that necessitates discussions on workforce and operational planning. There have already been examples of the risks that flawed or rushed AI implementations present. In our report, we outline a series of action items for effective and responsible public sector deployment, including publishing clear, plain-language strategies; equipping workers with AI tools, training and governance frameworks; introducing low-risk applications that pose smaller threats to workers; rolling job classification realignments to reflect job change; and longer-term workforce planning, undertaken with workers and labour unions.

In sum, the renewed AI strategy must tackle the issue of adoption from a three-pronged approach: helping non-adopters identify useful applications and trial AI use responsibly; assisting advanced adopters to optimize for longer-term productivity benefits; and supporting workers (and other impacted groups) through the disruption and transition that AI will bring. This should be coupled with longitudinal efforts to track AI intelligence around macro adoption trends, micro company-level best practices, productivity outcomes, workforce impacts and disruptions, and implications for policymaking. At the organization level, this should generate a body of evidence that documents both the modalities and factors associated with productivity-enhancing AI implementation.

Recommendations

1. Develop and promote AI adoption programming to leaders or organizations with low AI adoption to address knowledge gaps. This could involve scaling and adapting initiatives such as the Regional Artificial Intelligence Initiative (RAII) and AI Assist Program, and scaling programs that promote applied research projects between colleges and/or polytechnics and SMEs on topics related to adoption. 
2. Prioritize responsible AI adoption efforts in Canada’s nonprofit sector, with targeted investment support to scale and build on early sector-wide initiatives.
3. Establish data and analysis infrastructure to generate economy-wide AI intelligence on AI adoption, productivity, workforce and other key trends and impacts, through Statistics Canada, with academic and think tank partners. This should include extending and expanding critical analysis tools, such as the Survey of Digital Technology and Internet Use (SDTIU), to capture key firm-level factors (e.g. use of off-the-shelf AI solutions versus those built in-house).
4. Launch directed firm-level research to study, evaluate and share AI implementation best practices and productivity-enhancing applications in business and institutional settings. This could build on efforts by granting councils, Mitacs, and through partnerships with firms, academic researchers and think tanks.
5. Continue to address core digital modernization challenges, including setting clear, public AI strategies to transparently set goals and expectations; provide tools, training, and space for workers to experiment with AI; and foster proactive discussions with public sector unions about long-term workforce planning. 

---

THEME 2

Building Safe AI Systems and Strengthening Trust in AI

Public trust must be at the centre of Canada’s renewed AI strategy. Building public trust requires ensuring that Canadians feel confident that harms from AI will be promptly and effectively addressed if they arise. Trust is also essential for Canada’s AI innovators, and business and institutional adoption plans to succeed, as workers, clients and other stakeholders must have confidence that use of the technologies will not cause harm, worsen service quality, or be used as a careless excuse to eliminate jobs.

It is notable then, that Canadians’ trust in AI technologies and AI companies is low. Research by KPMG and Edelman find that Canadians rank among the most skeptical of AI worldwide. In the Dais’ 2025 Survey of Online Harms, OpenAI was among the least trusted companies to act in the Canadian public interest, rivalling social platforms Facebook and TikTok. This mistrust likely stems from harms Canadians experience with AI content online, including exposure to deepfakes, AI-enabled disinformation, and fraud. They are also concerned about irresponsible AI deployment, such as the rush by companies to release video generation tools that were immediately used to create harmful content, or reports of youth suicide and mental health crises exacerbated by AI chatbot use.

The primary narrative advanced by large AI companies and their supporters is that AI safety and regulation stifle innovation and limit growth, and that the companies have sufficient internal infrastructure to address user safety. However, our research shows self-governance is clearly insufficient. For example, in Human or AI?, we demonstrated that synthetic content labelling strategies (small unintrusive labels) were ineffective in informing users that content was AI-generated. In that same report, we demonstrated that effective approaches to flag synthetic content reduce engagement, creating a central conflict of interest for companies to self-govern.

Trusting companies to act responsibly without public input is not only ineffective, but inconsistent with Canadian interests, values of human rights, or public attitudes. Our survey found a strong majority of Canadians (nearly 70 percent) support government intervention to require tech platforms to act responsibly and reduce harmful content. In this section, we outline possible actions that the government can take to create pathways for Canadians to build trust in AI systems.

Effective AI Governance as the Basis for Public Trust-Building

While the government has signalled that it will not prioritize the re-introduction of a bespoke bill to establish a regulatory framework for AI, as previously attempted in the Artificial Intelligence and Data Act (AIDA), there are many other tools to establish AI governance. Regulatory clarity, coherence and international cooperation will encourage companies to adopt more AI and digital technologies, not less.

Effective AI governance requires strong and independent oversight, and clear accountability and enforcement mechanisms. Existing and proposed regulators and enforcement agencies (the Privacy Commissioner, the Competition Bureau, and proposed Online Safety Regulator) should be resourced to fulfill their investigative and enforcement mandates that new legislation would introduce. Further, to ensure coordination and consistent enforcement when AI systems impact multiple sectors with overlapping jurisdictions, we need to build dedicated technical capacity across departments and foster formal cooperation among regulators, such as through bodies like the Canadian Digital Regulators Forum.

The scope of AI systems cuts across multiple sectors, including health, environmental protection, heritage, finance, and national security, making it unrealistic for a single commissioner, agency, or law to manage all aspects of governance. To address the growing body of AI harms, the strategy should empower existing government institutions in assessing how current governance regimes apply to AI, as well as identifying areas where new measures are needed. The United Kingdom’s pro-innovation 2023 framework could be used as a model to enable a faster and coordinated pathway to AI regulation,by building on existing institutional capacities rather than creating entirely new structures. Building public trust in AI does not require novel or disruptive approaches; instead, it requires reimagining how existing systems can respond to our new environment.  

The government can address overlapping categories of harm through related legislative proposals. In the new proposed privacy and data protection bill, the latest effort to update the existing PIPEDA law after Bill C-27 (including the Consumer Privacy Protection Act and AIDA) failed to pass in the last parliament, could update privacy protections relating to private sector AI use. It should include provisions for regulatory enforcement of the new Children’s Privacy Code that the Privacy Commissioner of Canada is introducing, which could be essential in governing design and data capture from tools like AI chatbots geared to children and teens. A reintroduced Online Safety Act would also put requirements on social platform harms that AI is amplifying. 

Canada should consider where these legal frameworks align with international agreements, with a focus on working with values-aligned democracies. For instance, ratifying the Council of Europe Framework Convention on AI and Human Rights, Democracy and the Rule of Law, of which Canada is a signatory; assessing the impacts of the European Union’s AI Act as it takes effect; and engaging in multilateral processes like the G7 Hiroshima AI Process reporting framework, and other coordinated actions on AI. 

Canada should continue efforts to establish technical standards for AI and intersecting digital technologies. This includes both domestic standards, such as those developed by the Digital Governance Council, and the Standards Council of Canada's continued coordination of Canadian participation in international standards setting processes. Once developed, AI standards adoption could be required under legal frameworks (as Europe is doing), or common practice through procurement processes of governments and large enterprises (as many are doing for cybersecurity standards).

Finally, trust is built through the public processes of passing laws and setting guardrails. Previous efforts have failed in part because they lacked genuine public consultations. When the federal government tabled Bill C-27 in 2022, over 60 civil society organizations, corporations, unions and academics decried the lack of public consultation, including around the inclusion of AIDA with little warning. This was echoed at the Dais’ multi-stakeholder roundtable on the proposed AIDA bill. Similar concerns surfaced when this very AI consultation was announced, with the process criticized as unrepresentative, unnecessarily hurried, and favoring industry actors. Building public trust requires developing avenues for Canadians to inform and actively shape policy. 

A participatory democratic approach within Canada’s AI strategy could mean creating ongoing, meaningful channels for public involvement that extend beyond traditional consultations or surveys. This could include online and in-person deliberative assemblies, citizens’ panels, and community-based advisory boards that bring together diverse voices to debate trade-offs, set priorities, evaluate emerging risks and build consensus on values. It could also involve co-design processes with historically marginalized communities, participatory technology assessments, and transparent mechanisms for public feedback on AI deployment in sectors such as health, education, and public safety. 

Recommendations

6. Task existing government institutions with assessing how their respective regulatory regimes can be applied to AI and identify gaps, enabling a faster, coordinated, and adaptive regulatory approach built on existing systems.
7. Address overlapping categories of harms through related legislative proposals for privacy and online safety, with a focus on immediate challenges like deepfakes, child and youth protection, and limiting the spread of disinformation.
8. Empower new and existing regulatory bodies to oversee AI-related harms and risks under new and existing laws, and strengthen cross-regulatory coordination by creating formal mechanisms for collaboration among privacy, human rights, competition, and sectoral regulators to ensure accountable AI governance.
9. Prioritize international coordination of Canada’s AI governance policies and mechanisms under international agreements and with values-aligned allies like the European Union.
10. Support continued development of AI technical standards in Canada and internationally, and consider establishing requirements under Canadian legal regimes or through government and large enterprise procurement processes to encourage broader adoption, and facilitate technological and regulatory interoperability.
11. Institutionalize participatory governance as part of the new national AI strategy and its activities, which can include, but is not limited to, online and in-person citizens’ assemblies and community advisory boards, to give Canadians a real voice in shaping Canada’s approach to AI policy.

Building and Securing Canadian AI Sovereignty 

---

A recent Dais Commentary made the case that America’s AI Action Plan, released in Summer 2025, should be Canada’s wake-up call on AI and digital sovereignty. The US strategy asserts the goal of “achiev[ing] and maintain[ing] unquestioned and unchallenged global technological dominance,” while a section on international AI diplomacy and security commits to exporting the United States’ “full AI technology stack”—hardware, models, software, applications, standards—to countries in America’s so-called AI Alliance. Failure to go along means becoming a “rival.” 

In practical terms, this is about extending and solidifying the control that the US’ already-dominant big tech companies’ hold over the most transformative technologies of our time. This is a major threat to the economic security and digital sovereignty of Canada, Europe and other Western alliance nations, and the latest signal that Canada needs a bold, new national strategy for AI and the digital economy.

For Canada to make gains towards digital sovereignty through its AI strategy, we should take steps to reduce dependence on foreign technology and infrastructure. To be clear, it is not in Canada’s interest to pursue digital isolationism. However, our current pursuit of increased integration of AI in the economy risks Canada becoming beholden to foreign technology companies and governments. 

There are two pillars to secure Canadian sovereignty within the context of AI. The first involves identifying and safeguarding Canada’s unique contributions in the AI value chain. The second involves ensuring that there are Canadian alternatives for critical areas where AI technology is deployed, such as in defense and in health. 

The following section explores approaches to advance Canadian sovereignty in AI by cementing our place in the AI production infrastructure value chain, and by enabling Canadian companies to commercialize AI products.

---

THEME 5

Building Enabling Infrastructure

The first pillar to secure Canadian sovereignty involves identifying and investing in the value chain of AI-enabling infrastructure that ensures Canada’s national and economic security, and leverages Canada's unique competitive advantages. This means avoiding investing in areas where we lack clear competitive advantages (such as the design and manufacturing of advanced GPUs that a small number of companies in the US, Taiwan, and the Netherlands have largely dominated).

The Dais’ Can Canada Compute? study found that Canada lags behind all G7 countries in publicly owned compute resources. While it’s critical to address this aggregate compute capacity challenge, sovereign compute for strategic and security needs should be distinguished from general-purpose compute access for companies and organizations seeking to adopt and commercialize AI. The discussion here focuses on building sovereign AI compute (see the Dais’ AI compute consultation submission, as well as our discussion in Theme 6 of this submission for analysis on access).

An initial step must be clearly defining AI sovereignty for Canada. Global AI giants wield large power in controlling AI models and infrastructure, as well as the terms of AI deployment. OpenAI’s recent offer to assist in Canada’s sovereign AI tests the definition of sovereignty itself. It is therefore essential that the strategy introduce a clear and operationalizable definition of digital, cloud, and AI sovereignty. It should clarify how sovereign principles should apply to data governance, infrastructure ownership, or value chain control, among others, to guide AI strategy commitments and investments. 

With sovereignty defined, the strategy should then establish Canadian-owned and -operated AI computing capacity for Canada’s highest strategic priority AI applications (e.g. national and energy security), to ensure firms and public institutions are not independent from foreign providers and insulated from shifting geopolitics. These key strategic AI systems could be identified and developed through a Grand Challenges process (discussed in Theme 6). 

The strategy should direct identification of Canada’s strategic AI applications, and the creation of capabilities for forecasting the associated compute demands. There are a few pathways for building sovereign compute (including some detailed in our study, and actioned through the Budget 2024 Canadian Sovereign AI Compute Strategy). These include:

• Centralizing government procurement or public investment, subsidizing AI compute from existing domestic enterprise cloud computing providers under commercial requirements that assure sovereign ownership, control, and access (e.g. the federal government’s recent cloud procurement from ThinkON, and partnerships with Canadian  telecoms companies).
• Strategic international partnerships for shared AI compute infrastructure: jointly purchasing at scale, under assured access commitment, with values-aligned partners like the European Union, UK, Japan and South Korea (e.g. exploring participation in the European High-Performance Computing Joint Undertaking).
• Investment in domestic AI supercomputing infrastructure, expanding Canada’s network of 10 high-performance domestic AI supercomputing sites. 

At the same time, Canada holds a competitive advantage over many other nations (including the US) in our capacity to design, develop, manufacture, and generate reliable and sustainable energy sources that can address the immense power needs of data centres. From hydroelectric power, to nuclear, and comparatively more sustainable natural gas sources. This aligns closely to the government’s nation-building agenda, and allows for infrastructure that recognizes AI as a technology that encompasses modalities beyond generative AI, with vastly different energy and infrastructure needs.

Additionally, nation-building infrastructure investments should prioritize Canada’s AI sovereignty and security. The new Major Projects Office could play a central role in new AI and digital infrastructure projects. This should include public benefit AI investments (termed “public AI”), advanced through coalitions of public, private and nonprofit actors to build foundational technologies that spur an ecosystem of AI applications, both commercial and non-commercial, with civic purpose. Digital sovereignty through collaboration can pool resources across diverse groups, from artists to companies,  nonprofits, and alliance partners such as the EU. Existing models such as AI Commons and Public AI Network can serve as a guide.

In addition, investments in large-scale electricity projects to power AI infrastructure, such as new hydroelectric, expanded CANDU reactors, and small modular nuclear capacity that prioritize clean, reliable energy generation should be expanded. A forward-looking approach should centre energy sovereignty and sustainable infrastructure, ensuring that AI growth is powered by Canadian innovation and clean energy.

Absent a clear definition of sovereignty and guiding framework, Canada risks misaligning its AI strategy from security needs and comparative economic advantages, diverting resources toward unrealistic infrastructure goals, and deepening dependence on foreign solutions which undermine AI commercialization and adoption domestically. 

Recommendations

18. Develop and codify a formal definition of digital and AI sovereignty to inform the AI strategy, clarifying how digital infrastructure and protection of digital assets, and other aspects within the AI value chain require technological independence, and where there are opportunities for collaboration with allied countries. 
19. Focus resources in building sovereign compute capacity for Canada’s highest strategic priority applications, used to design, deploy and operate critical use cases (such as national defence), potentially identified through a Grand Challenge process (as explored in theme 6 on scaling AI). Three parallel pathways for building compute are identified (detailed in Can Canada Compute?).
20. Use the Major Projects Office to advance the national AI strategy, through both nation-building AI and digital infrastructure projects, and alignment to investments in clean, sustainable energy projects to cement Canada’s competitive edge in energy production that powers AI data centres.
21. Ensure public involvement, ownership, and international collaboration as a model for sovereign AI, modelled on existing efforts like Public AI, or the proposed AI Commons, by focusing on unique comparative advantages like ethical data, soft power and open source AI development.

---

THEME 6

Scaling Canadian Champions and Attracting Investment

The second pillar to protect Canadian AI sovereignty is by ensuring the existence of strong Canadian alternatives for key AI technologies. This depends on building a strategy that supports homegrown AI champions in critical usage areas, creates a supportive and equitable entrepreneurship ecosystem, expands access to the foundational infrastructure that underpins AI innovation, and addresses critical infrastructure gaps.

Efforts to help commercialize Canadian-made AI technology must focus on scale-ups, not just start-ups. Scale-ups are firms that grow quickly and have an outsized productivity contribution. Importantly, they are able to leverage research and development more effectively than smaller firms, and are more likely to export (see our previous work, Into the Scale-up Verse). Commercialization strategies must focus on company growth, not just creating start-ups.

A recent Dais convening (see From Potential to Performance) of industry experts revealed a prevailing view that existing innovation programming must be modernized with a focus on retaining top AI talent and intellectual property (IP). Canada's AI infrastructure and AI application start-ups and scale-ups needing globally competitive public support for IP creation, talent retention, and commercialization efforts. Existing innovation support programs such as the Scientific Research & Experimental Development tax credit, Industrial Research Assistance Program, the Canada Infrastructure Bank, the Strategic Innovation Fund, and Innovative Solutions Canada need tailored and responsive processes for Canadian AI enterprises. 

In addition to leveraging broad-based commercialization measures, the AI strategy needs to contend with how policies can help grow Canadian companies building different layers of the AI stack: infrastructure, data, foundational models, compute, and applications for use in critical areas (such as healthcare, security and defence, and supply chain management). As it is not feasible for Canada to excel in building every layer of the stack, Canada needs to identify critical and priority areas. The next AI strategy should clearly outline these priority usage areas with clear public-interest rationale, where no strong Canadian alternatives to foreign AI solutions exist. 

The government should launch a Grand Challenge program to identify strategic AI use cases that lack Canadian solutions and create matched contribution private-public investment funds to invest to build domestic capacity for those usage areas. Modelled on ideas like moonshot policy, a grand challenge could leverage existing incubator and accelerator infrastructure and programs like Lab2Market, Creative Destruction Lab, and others.

To ensure that AI industry development in Canada drives inclusive growth rather than deepening existing divides, policymakers should prioritize actions that expand shared infrastructure, close equity gaps in access and participation, and strengthen trust through responsible and transparent adoption.

Recommendations

22. Launch a Canadian AI Grand Challenge that clearly articulates areas where the government believes Canadian-made AI solutions must exist. The challenges should have clear, bold, measurable, and time-limited goals.
23. Launch matched private-public investment funds that invest in Canadian companies at all stages that are building solutions in critical AI usage areas without a clear Canadian alternative, as identified in a Canadian AI Grand Challenge.
24. Prioritize accessing public compute and data resources to companies that receive funding through Canadian AI Grand Challenge through the existing Canada Compute Access Fund.