When the Algorithm Becomes a Compliance Risk: What TA Teams Must Have Completed by August 2026 Under the EU AI Act
- Marcus
- 6 days ago
- 4 min read
The EU Artificial Intelligence Act (EU AI Act) has been in force since 1 August 2024. That part is widely known. What is far less understood: recruiting is one of the most heavily regulated domains under the regulation.
Most AI-driven hiring tools fall into the “high-risk” category. As of August 2026, extensive compliance obligations apply to any organisation using them. If you have not started preparing yet, you are effectively betting on time.
So what exactly must Talent Acquisition teams complete — and where should they begin?
What the EU AI Act Means for Recruiting — More Than Many Expect
The regulation classifies AI systems by risk level:
The Act is explicit: AI systems used for recruitment, applicant screening, and employment decisions are automatically classified as high risk, because they directly affect individuals’ access to employment.
This includes, among others:
Automated screening and ranking of applications
Chatbots that assess or pre-select candidates
Video interview platforms with automated evaluation
AI-based personality or competency assessments
Sourcing tools that prioritise or exclude candidates
The scope is broader than many TA leaders initially assume.
According to Littler’s 2025 European Employer Survey, fewer than 20% of companies report feeling “very well-prepared.” Most have neither created a complete inventory of their HR AI tools nor conducted a formal risk classification.
And the regulation is not limited to European companies. It applies whenever AI outputs are used for EU-based candidates or employees — even if the employer is headquartered in the US or Asia and has no EU subsidiary.
The Three Deadlines — And What They Really Mean
The EU AI Act does not become fully applicable overnight. It rolls out in phases, which creates a false sense of security.
2 February 2025 – Already Passed
From this date onward, certain AI practices are prohibited, including:
Emotion recognition in job interviews
Biometric categorisation based on ethnicity or other protected characteristics
AI systems that manipulate candidates psychologically
If your organisation still uses video interview platforms to analyse facial expressions, tone of voice, or gaze direction for evaluation purposes, this deadline has already been missed. Immediate action is required.
2 August 2025 – Primarily for Providers, Indirectly for Users
Providers of general-purpose AI models embedded in HR tools must document training data and publish transparency reports.
TA teams may not be directly responsible, but vendor non-compliance does not shield the user organisation. Vendor due diligence becomes essential.
2 August 2026 – The Core Deadline for Employers
From this date, full high-risk compliance requirements apply to all deployers:
Technical documentation
Demonstrable bias testing
Established logging mechanisms
Human oversight
Active candidate information
Fines can reach €35 million or 7% of global annual turnover, whichever is higher, for serious violations.
Six months may sound manageable. In reality, inventorying systems, conducting risk assessments, negotiating with vendors, adapting processes, and training teams — all in coordination with Legal, IT, and Compliance — requires substantial lead time.
What Must Be Switched Off or Reviewed Immediately
The following practices have been prohibited since February 2025:
Emotion detection in video interviews
Biometric categorisation by gender, ethnicity, or similar attributes
Social scoring based on online behaviour or reputation data
Psychological profiling unrelated to role-specific behaviour
Many vendors have removed such features. Not all have. Review product documentation and obtain written confirmation of compliance. Verbal assurances are insufficient.
The Practical Checklist: Seven Steps to Start Now
A pragmatic entry point for TA teams:
Inventory your AI systems
List all AI tools used in recruiting — including components embedded in ATS, HRIS, sourcing platforms, or video tools. Many organisations use AI without having consciously selected it.
Classify risk levels
If the tool is used for screening, ranking, or candidate evaluation, assume it poses a high risk. Verify whether prohibited features are active.
Conduct vendor audits
Request written confirmation that systems are — or will be by August 2026 — compliant with the EU AI Act.
Ensure human oversight
High-risk AI may not make fully automated final decisions. Human review must be meaningful and documented.
Inform candidates transparently
Candidates must be informed when AI is involved in decision-making. Update privacy notices, application processes, and invitation communications accordingly.
Implement logging mechanisms
Logs must be retained for at least six months. Confirm technical feasibility with IT.
Train your teams (AI literacy)
The Act explicitly requires AI literacy. Recruiters and HR decision-makers must understand the systems they use.
Talent Acquisition Cannot Delegate This.
The EU AI Act is not an IT-only or Legal-only topic. It directly affects the core of modern recruiting operating models — automation, scalability, and data-driven decision support.
Compliance is not anti-innovation. But TA leaders must understand which decisions their systems prepare for or influence—and be able to justify them.
Interestingly, companies that begin with inventorying their tech stack often discover something uncomfortable: they know their own AI landscape less well than they thought. AI components are frequently embedded in tools that are never consciously chosen as “AI systems.”
Fewer than 20% of European employers consider themselves well prepared. For TA leaders who act now, this is not just risk mitigation — it is an opportunity.
Compliance, done properly, creates process clarity. And that clarity becomes the foundation for using AI in recruiting responsibly and strategically beyond 2026.
Sources
Ogletree Deakins, “The EU AI Act Is Here – What It Means for U.S. Employers” (October 2025)
Littler, “European Employer Survey 2025” (December 2025)
EU AI Office, “Prohibited AI Practices & Timeline”
https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai
HireTruffle, “Recruiting under the EU AI Act – Full Guide 2025”
Boundless HQ, “What is the EU AI Act?” (December 2025)
https://boundlesshq.com/blog/what-is-the-eu-ai-act-everything-you-need-to-know/
Greenberg Traurig, “Use of AI in Recruitment and Hiring – EU and US” (2025)
Comments