In today’s fast-moving AI landscape, one of the biggest risks isn’t what you can see—it’s what you don’t. As organizations outsource, partner, buy in AI systems, or build vendor dependencies that embed AI, the question of visibility becomes central: who has access to what data, who controls the AI, how transparent is the process, and what happens when something goes wrong.
A recent Kiteworks survey of 461 organizations across North America, Europe, APAC and the Middle East underscores this issue:
- Nearly half of organizations don’t know how many third parties they’re dealing with, or how often data breaches happen.
- Only 17% have fully implemented technical AI governance controls.
- And 83% lack automated controls to prevent sensitive data from being shared with public AI tools.
This is more than a technical gap. It’s a governance gap — and HR sits right in the middle of it.
EU AI Act: why HR cannot ignore third-party AI
The EU AI Act entered into force in August 2024, creating the world’s first comprehensive legal framework for AI. Its obligations for high-risk systems — including AI used in recruitment, performance management, and other core HR functions — are approaching fast.
That deadline matters. High-risk AI systems will need documented transparency, traceability, human oversight, and rigorous data quality standards. Importantly, organizations deploying AI in HR will remain accountable even if the AI is provided by a third-party vendor. The law makes it clear: “we didn’t build it” is not an excuse.
ISO 42001: turning regulation into practice
ISO/IEC 42001, the standard for Artificial Intelligence Management Systems (AIMS), provides the practical scaffolding to meet these obligations. Among its most relevant requirements for HR:
- Third-party and supplier oversight is mandatory — organizations must evaluate risks not just in their own AI systems but also in those of vendors and subcontractors.
- Continuous monitoring of AI systems across their lifecycle — from design to deployment to updates — ensures that risks don’t creep in unnoticed.
- Competence and literacy requirements — staff must be trained to understand AI’s risks and limitations, and to intervene when necessary.
Adopting ISO 42001 is not a legal requirement, but it is one of the clearest ways to show regulators, employees, and partners that your organization is serious about responsible AI.
The questions HR leaders must start asking
So what does all this mean for HR in practice? It means moving from being a passive consumer of vendor tools to an active steward of AI governance. That starts with asking the right questions:
- What AI systems are we actually using in HR? From ATS and screening platforms to performance analytics dashboards — do you have an inventory of where AI touches employee data and decisions?
- How do these systems make decisions? Vendors may be reluctant to share full model details, but HR must insist on at least a functional explanation: what data is used, what criteria matter, and how bias is addressed.
- What contractual obligations bind our vendors? Do agreements require them to keep logs, provide documentation, permit audits, and maintain human oversight? Without these, compliance is a guesswork exercise.
- Who owns the risk inside our company? If a vendor’s AI discriminates, leaks data, or malfunctions, who in HR is accountable? Clear ownership is essential to avoid paralysis when something goes wrong.
- Are our people trained to challenge AI? Professional skepticism is not optional. HR staff and managers must know how to question outputs, identify red flags, and intervene effectively.
- Do we have an incident plan? When an AI system fails, whether in-house or vendor-provided, is there a clear process for escalation, communication, and correction?
Asking these questions doesn’t just reduce legal and reputational risk. It also builds trust with employees — who increasingly expect that the AI tools evaluating them are fair, transparent, and well-governed.
Conclusion: from blind spots to clear sight
Third-party AI risk isn’t a niche concern. The Kiteworks data shows that most organizations are still flying blind. The EU AI Act gives legal teeth to the need for visibility and oversight. ISO 42001 offers a structured way to operationalize those requirements.
For HR leaders, the message is simple: don’t wait for the deadline to act. Start by asking the hard questions now. Build visibility, demand accountability from vendors, and ensure your teams are trained to be more than passive users — they must become active, educated skeptics who can steer AI rather than be steered by it.
Because in the AI era, ignorance is not just a weakness. It’s a liability.
Vendor AI is still your accountability. Build the management system that proves oversight.