AI Risks in HR: Privacy, Bias and a Lack of Strategy
What are the main risks and barriers of AI in HR?
According to the 1st Barometer of AI in HR in Canada (NexaRH and Parlons RH, 2026), confidentiality and privacy top the list of risks, ahead of discrimination and bias. Yet the main barrier to deployment is not technological: 76% of HR professionals cite the lack of a clear strategy.
Key figures on AI risks in HR
32% of respondents cite confidentiality and privacy as the first risk factor, and the most serious.
25% mention the risk of discrimination and of reproducing assessment and decision biases.
17% highlight the multiplication of uses and tools without overall coherence.
76% name the lack of a clear strategy as the main barrier to AI adoption.
47% cite a lack of skills as the second barrier and 43% budget constraints.
19% mention the lack of demonstrated return on investment.
Not surprisingly, confidentiality preservation and the protection of personal information come at the top of the list of risks, closely followed by the reduction of assessment and decision-making biases.
Naturally sensitive to these issues, HR are also the primary legal custodian of sensitive employee data, since the enactment of Canadian and Quebec laws inspired by the European Union’s General Data Protection Regulation (GDPR).
4.1 The most serious risks
With 32% risk severity, confidentiality compliance outweighs all other risks. Since the launch of ChatGPT in December 2022, followed by its competitors, concerns have arisen regarding the use of confidential data for training language models (LLMs). The paid versions, which are more secure, continue to be scrutinized by security and IT teams tasked with thoroughly reading contract terms and conditions and decoding technological architecture and data hosting.
Secondly, assessment and decision-making biases are viewed as a high risk by 25% of respondents. These biases are inherent in any algorithmic recommendation or prediction model and all language models (LLMs) and must be considered when selecting tools and using them for recruitment, training, or predictive analysis. Everything that comes from AI remains a human responsibility. Therefore, it is necessary to verify, understand, review, and validate according to the Human in The Loop principle. The more advanced and integrated the intelligence, the greater the accountability of decision-makers.
Employees, who often engage excessively with their pocket AI, represent a major source of data leaks. They must understand the risks and limits of individual solutions used at work while waiting for approved AI solutions to be integrated according to best practices.
Finally, the third risk factor is the lack of coherence among AI tools. With 17% risk severity, HR professionals sense that the omnipresence of AI in solution offerings poses a challenge. Just as they are beginning to familiarize themselves with AI applications at work, the abundance of tools calls for priorities, choices, and clarity.
4.2 Change management: a blind spot?
It is noteworthy, consistent with all other indicators, that resistance to change does not emerge as a high-risk factor. The use of generative AI, even if not always intensive, is already in many hands.
Unlike technological changes driven by organizations, generative AI has entered from the bottom up. However, one can expect that Agentic AI, requiring orchestration between enterprise solutions and external solutions, will be more disruptive in nature, requiring more sustained design and organizational change management (OCM).
4.3 A call for clarity!
The barriers to deploying AI in HR are clearly identified by 76% of respondents who say that a clear strategy is lacking.
The absence of an AI roadmap in HR indicates two things: HR leadership initially followed operational teams regarding AI, without feeling concerned by these new technological capabilities. Initially cautious about taking a stance on these issues, HR decision-makers must now regain control of AI for their own destiny and the benefit of their professionals.
The lack of internal skills is the second barrier cited by 47% of respondents. Skills development within teams must be achieved through training and development. While budgets are expected to be maintained overall for one-third of respondents or increase, for one-quarter of HR departments, AI skills become a capacity issue, especially in the context of labour scarcity and the rapid emergence of new skills. The ability to recruit and train is crucial to the success of an AI-First strategy.
Budget constraints represent the third barrier. 43% of respondents state that budgets limit projects. Nevertheless, return on investment ranks near the bottom among organizational barriers to AI in HR. It is evident here that HR respondents do not have control over investments. AI in HR is not a priority for decision-makers, who see it more as part of the continuous modernization of work tools than as a project focus.
Interview: Nicolas Bonnafous, Partner, Innoova
“An essential tool for modernizing HR practices.”
How to transform the budgetary constraint of AI in HR into a strategic lever?
Artificial intelligence has now become an essential lever for modernizing HR practices. Yet, on the ground, this momentum is often slowed by limited budgets and small technical teams. Moving forward despite these constraints requires a pragmatic approach: securing quick wins, developing a truly transformative vision, and effectively managing change. The ultimate goal remains the same: integrating AI without ever losing sight of its impact on employees’ daily lives.
Focus on Quick Wins When Resources Are Limited.
When resources are tight, I start by activating AI features already built into existing tools: writing assistants, scheduling aids, and so on. These functions boost efficiency without additional investment and keep complexity low.
Next, I target repetitive administrative tasks: sorting résumés, scheduling interviews, answering common questions. These activities are ideal for automation through simple solutions. Generative AI also helps with drafting job descriptions, preparing HR communications, and summarizing annual reviews. These “quick wins” provide tangible proof of AI’s value and serve as a springboard for more ambitious projects.
Moving from Tactical to Transformational.
True HR modernization isn’t just about doing things faster. It’s about reimagining how we work, collaborate, and interact. I aim for AI adoption that enriches the experience for all users.
To make this transformation sustainable, I ensure clear governance of tools, transparency in how they operate, and explanations of the logic behind their recommendations. This educational approach builds trust and makes integration easier.
Every AI initiative is aligned with concrete business priorities: reducing time-to-fill for key roles, improving retention, and anticipating future skill needs. AI in HR is not a tech project: it’s a strategic initiative serving organizational performance.
Putting People at the Centre of Change Management
Technology is only part of the equation: adoption depends first on the human factor. I work to reduce fears of replacement by clearly positioning AI as a tool to augment human capabilities. AI processes information: HR professionals and managers retain the roles of judgment, relationship-building, and empathy.
I move forward gradually: pilots on limited scopes, clear objectives, precise indicators, and successive iterations. This approach allows us to adjust usage, secure gains, and ensure stakeholder buy-in. Moving too fast can create lasting resistance; advancing methodically fosters ownership.
Conclusion
AI in HR is not an end in itself, it’s an accelerator of relevance and transformation. By combining quick wins, long-term vision, and human-centered change management, it becomes possible to turn budget constraints into a strategic advantage. The result: an HR function that is more agile, more predictive, and… even more human.
4.4 Implementing AI: the task will be challenging
Although HR respondents express confidence and enthusiasm, they also report that implementing AI is rather difficult (57%), or even very difficult (7%).
The difficulty of implementing AI in HR is proportional to company size. In large enterprises (LEs), respondents find it rather difficult (65%) or very difficult (9%), whereas in small enterprises (SEs), HR professionals estimate the level of difficulty at 50% and 4%, respectively. Overall, the weight of governance and reduced agility in large enterprises act both as a means of protection and a brake on adoption.
Although exposed to new risks, small and medium-sized enterprises can move faster with these disruptive technologies.
Expert insights: Cynthia Chassigneux, Lawyer, CHX Avocat Inc.
In human resources, as in any other sector, the installation and use of artificial intelligence ("AI") tools or systems by employees without authorization or validation from technical and legal teams raises many issues for employers. The absence of a governance framework exposes the company to significant risks.
Among the issues posed by AI, particularly by Shadow AI, we can think of the security and integrity of systems, the unauthorized disclosure of confidential data (e.g., intellectual property, business strategy), or personal information ("PI").
The processing of PI of job applicants or employees by AI solutions must be considered, among other things, in terms of transparency, security, and confidentiality. This requires informing individuals when a decision based exclusively on automated processing of their PI is made about them. This presupposes conducting a privacy impact assessment (analysis of risks and of the level of protection granted to PI if cross-border flows, mitigation strategy) before deploying such solutions.
These considerations are not new. A legal and technical framework exists to regulate the processing of PI, regardless of the medium. However, awareness, the adoption of internal standards regarding the permitted and prohibited uses of AI, and monitoring are and remain the key words, otherwise the company's liability could be engaged.
Expert insights: Pierrick Plusquellec, Associate Professor, Université de Montréal, Scientific Director and Industry Collaboration, Obvia
“The findings are clear: HR departments are largely optimistic about AI.”
More than 91% of respondents see it as an opportunity and say they are confident about the future of their organization. However, this optimism is not yet backed up by concrete action: less than 20% of organizations have dedicated governance or funding for AI.
The top three perceived benefits illustrate this tension well. At the top of the list are gains in efficiency, productivity, and time. But a key question remains: what do we do with this freed-up time? Next comes the automation of low-value-added tasks, which requires careful consideration of the skills we are willing to delegate to AI, those that remain essential, and those that urgently need to be strengthened.
Finally, strategic refocusing on people comes in third place. This is a strong signal, but still vague. What does a “more strategic” HR role mean in concrete terms? And how can managers be equipped to take on this role?
Conversely, the bottom 3 highlights very real vulnerabilities: feelings of being overwhelmed, low AI literacy, and concerns about the environmental impact and responsible use of these technologies. These issues are even more concerning given that few organizations, especially smaller ones, have put in place safety and ethical frameworks to protect individuals and the organization.
It is important to remember that well-being at work is not limited to the absence of risks. It refers to the feeling of functioning well, feeling competent, engaged, and supported, and to the daily emotional experience (Diener et al., 2009). This calls for training efforts, not only in AI, but also in emotional intelligence.
Finally, while future risks related to confidentiality, bias, and discrimination have been clearly identified, one blind spot remains: the loss of autonomy and expertise, as well as the illusion of competence among the next generation.
It is therefore urgent to shift the conversation. Beyond the tools themselves, AI raises above all a question of skills: those to be delegated to systems and those to be cultivated in humans.
The 7 parts of the AI in HR Barometer 2026
Chapter 4. Risks and limitations of AI in HR (you are here)
Frequently asked questions
What is the main risk of AI in HR?
Confidentiality and privacy: 32% of respondents to the 2026 AI in HR Barometer rank it first, and 91% are concerned about it. Discrimination and the reproduction of bias come next (25%).
What is the main barrier to deploying AI in HR?
The lack of a clear strategy, cited by 76% of respondents, followed by a lack of skills (47%) and budget constraints (43%).
Is it difficult to deploy AI in the HR function?
Yes for 64% of respondents, who find it rather difficult (57%) or very difficult (7%). Difficulty grows with organization size: 74% in large enterprises versus 54% in small ones.
Source and methodology
Source: 1st Barometer of AI in HR in Canada, 2026 edition, published by NexaRH in collaboration with Parlons RH. Pan-Canadian online survey conducted from October 1 to November 30, 2025 among 336 respondents (80% HR professionals within organizations, 20% HR service providers), in French and English, complemented by a discussion workshop with HR vice-presidents on November 12, 2025.
Read the full overview: 1st Barometer of AI in HR in Canada 2026.
Next chapter: Is the HR function immune to transformation?.
How to cite this barometer
NexaRH and Parlons RH (2026). 1st Barometer of AI in HR in Canada, 2026 edition. Survey of 336 HR professionals, October and November 2025. Available at: www.nexarh.com/en/post/ai-hr-barometer-canada-2026
Comments