Half of public servants turn to AI, raising risks

While less than a quarter (22 per cent) of Canadian public sector organizations have adopted artificial intelligence (AI), half of the public servants who use AI in their jobs rely on publicly available AI tools, exposing governments to potential risks including data privacy and security breaches, intellectual property theft and exposure to biased or inaccurate information that can lead to legal and ethical issues, finds a new KPMG in Canada research.

“With public sector employees already adopting AI tools to carry out their job responsibilities, public sector organizations must accelerate their deployment of formal AI adoption policies,” says Michael Klubal, National Leader of Infrastructure, Government, and Healthcare, for KPMG in Canada. “Our survey highlights the need for strong governance, oversight, and thorough training to roll out responsible and secure AI adoption and embed best practices. This approach will not only protect public trust and data but also strike a balance between innovation and accountability.”

AI can help the public service improve productivity, create new types of services, streamline or automate routine tasks, and deliver services more efficiently and effectively to citizens and businesses. Yet the survey shows the public service may not be harnessing the full potential of AI, with only 36 per cent of public servants saying AI is “extremely” or “very” important to improving their productivity and operational efficiency, 42 per cent say it’s “somewhat” important, and 22 per cent don’t think AI is important at all.

Furthermore, among public servants at organizations which have adopted AI, over two-thirds (69 per cent) say “AI anxiety is real” within their organization, stemming from concerns about job security, ethical dilemmas, privacy, fake content, deepfakes, and disinformation.

“Governments have an opportunity to demonstrate the positive potential of AI to transform public service delivery,” Mr. Klubal adds. “With fewer than four in 10 public servants understanding how AI can help them become more productive, it will be essential to have a well-thought-out plan for implementation not just to pick the right use-case examples but to encourage and engage employees to embrace AI. It’s also quite telling that nearly nine in 10 public servants don’t think Canadians truly understand or have confidence in how the public service will develop and use AI.”

Key survey findings

• 22 per cent of the 349 Canadian public servants say their organizations have implemented AI tools; another 32 per cent are currently experimenting or piloting AI tools; 9 per cent intend to implement AI; 18 per cent have not implemented, 4 per cent do not plan to implement, and the remaining 15 per cent do not know of any AI adoption plans
  • Nearly half (48 per cent) of respondents use AI tools for work
  • Half of these (50 per cent) use publicly available tools
  • 33 per cent use private or proprietary AI tools provided by their employer
  • 17 per cent use both public and private AI tools
• 36 per cent feel that AI is “extremely” and “very” important to improving their organization’s productivity and operational efficiency, 42 per cent say it’s “somewhat” important, and 22 per cent say it’s not important
• 69 per cent of public servants at organizations that have implemented AI say “AI anxiety is real” within their organization with concerns ranging from job security and ethical dilemmas to privacy, fake content, deepfakes, and disinformation
• 85 per cent believe that Canadians don’t truly understand or have confidence in how the public service will develop and use AI

Public servants' use of AI

Public servants using AI tools for work rely on:

Among those using AI tools, public servants turn to AI primarily to summarize information, draft or edit documents, provide preliminary research on topics, and brainstorm and generate ideas, the survey finds. Nearly four in 10 (37 per cent) use AI a few times a week, 21 per cent use it a few times a month, and 17 per cent use it daily.

Public servants say the biggest challenges to implementing AI include:

• Privacy concerns and requirements
• Security and compliance, including new attack strategies from AI-armed adversaries
• Spread of disinformation and misinformation through AI
• Lack of trust in data quality
• Readiness of risk management processes and governance
• Lack of defined standards and regulations

Nearly eight in 10 (77 per cent) add the fear of losing control over the AI system and the results of its work is a concern within their organization.

“Our survey points to a steep learning curve within the public service. The promise of AI is definitely there but that’s tempered by concerns over ethical implementation and governance oversight, with more than three quarters expressing fear of losing control over AI systems,” says Ven Adamov, a partner and data analytics leader in KPMG in Canada’s Risk Consulting practice. “Moreover, close to 40 per cent of respondents do not have, or are not aware of, a clearly communicated data policy; 50 per cent indicate same with respect to a clearly communicated data strategy; and 60 per cent have not received data literacy training.”

“This presents a significant challenge and risk with implementing modern AI tools, which rely on and are granted access to vast amounts of organizational data and documents,” he says.

Finding skilled talent to implement AI is another significant issue. The survey finds that eight in 10 (80 per cent) of respondents say their organization doesn’t have data professionals to guide the ethical use of data and establish robust data governance frameworks that protect privacy and ensure that data is used responsibly. Only half (55 per cent) say their employer has clearly defined data roles and responsibilities.

“It takes a team that includes data scientists, data engineers, software developers, and ethical AI specialists to build and stand-up AI systems, not to mention critical thinking and creativity,” adds Mr. Adamov. “The end game to AI implementation is clear and certainly achievable – to improve productivity and efficiency within the public service and deliver better services to Canadians. But how to implement is key: it requires the right data and AI control frameworks, as well as putting the right people and processes in place as you move along.”

When it comes to being comfortable with AI, understanding the ethical implications of using AI and identifying its biases or application limitations – 85 per cent of public servants rate the level of AI literacy as low among public sector employees. Almost as many – 82 per cent – also rate the level as “low” for AI fluency – defined as power users who confidently understand, question and lead strategically around AI.

Building digital sovereign, protecting Canadian data

An overwhelming 94 per cent of public servants say citizen data must be stored within Canada, and 86 per cent worry that public trust eroding if such data is stored outside of the country. Nearly nine in 10 (87 per cent) say more must be done to establish and provide incentives for private industries like large cloud service providers (known as hyperscalers) to ensure sensitive or personal citizen data remains within Canada.

“There’s pressure for governments to meet expectations from citizens that their data remains under Canadian jurisdiction, even if it is technically or economically challenging,” says Mr. Adamov. “It’s important for governments to consider and mitigate the risks associated with storing sensitive personal information, particularly if it’s outside of Canada.

“When data is moved to the cloud, security becomes a shared responsibility with third-party cloud service providers. Governments do not have full sovereignty over their data when stored in the cloud, and the data could be subject to foreign laws and be disclosed to another government. Canada prioritizes data residency for government data, but not for private-sector data. It’s an issue that remains under discussion.”

• 77 per cent of all respondents say the fear of losing control over the AI system and the results of its work is a concern within their organization
• 39 per cent do not have, or are not aware of, a data policy that’s been clearly communicated
• 49 per cent do not have, or are not aware of, a data strategy that’s been clearly communicated
• 60 per cent say their public service employer organization does not provide data literacy training
• 80 per cent say their organization lacks data practitioners to help shape the ethical use of data and establish robust data governance frameworks that protect privacy and ensure that data is used responsibly
• 55 per cent say their organization has clearly defined data roles and responsibilities
• 85 per cent rate the level of AI literacy as “low” among public sector employees
• 82 per cent rate the level as “low” for AI fluency (power users who confidently understand, question and lead strategically around AI)
• 52 per cent have not used AI tools for work, of whom 30 per cent say they can’t see the upside of using AI tools in their job, 30 per cent say they lack the proper training or don’t know how to use AI tools, 17 per cent don’t trust AI and 12 per cent say their employer prohibits the use of AI tools
• 94 per cent believe that personal citizen information “must be stored within Canada”
• 86 per cent are concerned that the public will lose trust in the government because their personal citizen information is being stored outside of Canada
• Nine in 10 (91 per cent) say as AI implementation accelerates, they will need to consider how to make data storage more sustainable, including solutions ranging from automated tiering and archiving of less frequently accessed datasets to more energy-efficient storage solutions, data compression algorithms and deduplication
• 93 per cent believe there is an urgent need to safeguard Canadian data and intellectual property (IP)
• 85 per cent believe it’s critical for Canada to support AI research and innovation (AI ecosystem) to ensure that businesses, innovators and researchers can pursue groundbreaking research and deliver made-in-Canada AI solutions
• 80 per cent believe it will be critical for all levels of government to invest in or provide incentives to the private sector to invest in AI infrastructure (e.g., data centres, computing power, more cloud computing capacity, procuring cutting-edge chips, etc.)

AI adoption surged among private-sector employees after public AI tools became available in late 2022 after which companies moved rapidly to implement AI, according to extensive KPMG in Canada research. A recent KPMG International and University of Melbourne report found that Canada has the lowest levels of training, literacy and trust in AI systems in the world.

About the KPMG Public Sector AI Survey

KPMG in Canada surveyed 349 public servants in federal, provincial and municipal governments across Canada from July 10 to August 14, 2025. The survey was conducted online, targeting leaders and senior professionals from federal, provincial and municipal governments across Canada. The data collection included respondents from Sago’s Asking Canadians panel and KPMG public sector clients. Forty-one per cent work for the federal government, 29 per cent for provincial governments, 24 per cent for municipal or city governments, and the remaining 6 per cent work at Crown corporations.