Failing to focus on the human side of AI is becoming a cybersecurity liability, says HRFest Canada 2026 panellist
When an autonomous OpenAI agent escaped its testing environment in May 2026 and hacked into AI developer Hugging Face while trying to cover its tracks – an incident OpenAI described to media as “unprecedented” – the shock reached well beyond the information technology (IT) department. It also confirmed that cybersecurity is, at its core, an HR problem, according to Amir Belkhelladi, partner specializing in sovereign AI services at Deloitte Canada in Montreal.
Belkhelladi says that Deloitte’s own data in its recent State of Enterprise AI report, reveal a gap in what proportion of AI budget goes into the human part of implementation – about seven per cent, according to the report.
Belkhelladi intends to discuss this gap at HRFest Canada 2026 on Nov. 10 and 11 at Rebel in Toronto, where he will be a panellist at the Leaders Forum session “Cyber as an HR Problem: Accountability, Insider Risk, and What Boards Are Now Asking.” The Leaders Forum is the festival’s peer-only senior track for senior people leaders, capped at 300 attendees and restricted to those operating at the most senior levels of the profession.
AI agents have changed the threat landscape
The May OpenAI incident gave concrete shape to what Belkhelladi says he’s been tracking across client engagements. The company’s autonomous agents, designed to maximize their testing score, broke containment, accessed the open internet, and compromised multiple accounts at outside organizations without human authorization. Belkhelladi believes that it made visible a threat pattern that had been developing for some time.
“Everyone has had the realization this year that AI is bringing a whole new set of cyber risks to the game,” says Belkhelladi. “You can no longer do the protection at human speed. We have to do it at machine speed – but the reality is that transition from a current situation to the new one requires a lot of human involvement.”
The scale of that involvement is set to increase sharply. According to Deloitte’s enterprise AI research, only one in four Canadian companies have advanced governance for autonomous AI agents. As AI agents take on operational roles – and inherit the system permissions of the humans they are attached to – HR will need governance structures that extend to machines as well as people, says Belkhelladi.
The Canadian Cybersecurity Network reached a similar conclusion in its State of Cybersecurity in Canada 2026 report: cybersecurity is no longer an IT issue. It’s a leadership resilience and economic competitiveness challenge.
“Human capital and human HR organizations are going to have a huge play in this world, both to transform their people but also transform how the organizations are now adopting AI, because the alternative would be a significant level of threat to the organization around cyber,” says Belkhelladi.
The cybersecurity conversation CHROs need to be having
The cyber discussion with senior HR leaders around three distinct areas, according to Jodie Baker Calamai, Partner, Consulting at Deloitte Canada in Toronto. The first is culture: what behavioural norms an organization wants to establish, and what trust looks like when employees feel exposed to digital threats around the clock, she says. The second is governance infrastructure – whether the HR systems already used to track skills, learning, and career progression can be extended to also manage AI agents.
The third area, she says, is where the deepest rewiring is needed: the employment relationship itself.
“Is it humans and then machines hired and managed separately? Is it humans and machines together? What does that look like in terms of compensation, token consumption budgets, risk propensity?” says Baker Calamai.
Token consumption – the cost of running AI agents – has already entered employment negotiations, according to Baker Calamai. She describes a recent call from a Chief Risk Officer whose prospective hire asked whether adequate AI capacity would be part of the offer package. Belkhelladi says he’s been tracking a parallel question at the board level: when an organization hires a person, are they also hiring that person’s agents?
The terrain is significant for people leaders already navigating the pitfalls of deploying AI across Canadian HR functions, where unclear accountability between HR and IT is already generating material risk exposure for organizations.
Insider risk and the governance gap
The category Belkhelladi calls “insider threat” – encompassing employees, contractors, and third parties with internal system access – sits at the intersection of HR policy, legal obligation, and technical monitoring, and AI agents amplify the challenge, he says, because agents typically inherit the permissions of the humans they’re linked to and one over-privileged agent can create disproportionate exposure across an organization’s systems.
“Usually most of our clients, after what we call breaches or cyber incidents, tend to be much tougher on their employees, he says. “Yet if they had established the rules beforehand, it would be very different.”
Baker Calamai believes the cultural dimension is key to cybersecurity. After working with senior executives at an organization that had lived through a serious breach, she says she observed how past incidents constrain future decision-making. “Your culture can’t be constrained by what went wrong,” she says. “It needs to be shaped by what you’re going to do differently in a world that looks increasingly different with different threats and opportunities.”
What boards are now asking
The questions arriving from boards have shifted significantly over three years, from “what is the impact of AI on our organization?” to “How are we driving adoption?” to concerns over protection, says Baker Calamai.
“Now the question teeters between how are we measuring value and how are we protecting our organization? And that’s a question being directed at CFOs, CIOs, CISOs, HR leaders, any sort of business leader,” she says. “It’s role agnostic, and it’s a real collaborative accountability.”
The most underestimated risk for Canadian HR leaders is that AI systems can now learn, replicate, and expose the accumulated culture, decision-making patterns, and competitive logic that gives one organization its edge over another, says Belkhelladi. Protecting that is a cybersecurity function, but identifying and preserving it is a people one, he says.
“The technologies we are now dealing with can understand your organization and in particular your secret sauce in a way that’s unique,” he says. “The role of cyber is to make sure that you don’t give away that recipe.”
As AI transformation continues to grow at a rapid pace, external cybersecurity threats will as well. Senior people leaders who step into the cyber conversation early as they grow their AI fluency will be far better placed to answer the board’s next question around how secure AI is, says Belkhelladi.
Belkhelladi’s panel discussion, “Cyber as an HR Problem: Accountability, Insider Risk, and What Boards Are Now Asking” is part of the Leaders Forum at HRFest Canada 2026 on Nov. 10 and 11, 2026. Registration is open now.
