If you ever watched Little Britain, you’ll remember the sketch where the unhelpful clerk, without a trace of emotion or logic, rejects every request with a deadpan “Computer says no.” I love that show. It is absolutely hilarious.
AI-driven CV screening tools now make the first cut in thousands of hiring processes — deciding who moves forward and who never even reaches a human recruiter.
The promise of automation is efficiency. But when that “no” is issued by an algorithm — without human oversight, transparency, or a chance to contest — it raises serious questions about fairness, accountability, and compliance with GDPR and the EU AI Act.
1. The GDPR view: automated decision-making
Article 22 GDPR gives every individual the right not to be subject to a decision based solely on automated processing that has a significant effect on them.
Being excluded from a job opportunity because an algorithm decided your CV doesn’t fit clearly meets both parts of this definition:
- it’s solely automated (no human review before rejection), and
- it has a significant effect (loss of a real-world opportunity).
So yes — in most cases, AI-based pre-screening is automated decision-making.
2. What regulators say
European regulators have been consistent:
- EDPB (WP29): pre-selection or automatic exclusion is automated decision-making.
- CNIL (France): human involvement must be real and not symbolic.
- BfDI (Germany): requires explicit consent or another lawful basis plus safeguards.
- Datatilsynet (Norway): even “decision-support” tools can breach Article 22 if the recruiter just accepts the AI’s recommendation.
In other words, if no meaningful human review happens before rejection, it’s automated decision-making — period.
3. The “human in the loop” myth
Many vendors and deployers argue: “A human makes the final decision — so the process isn’t automated.”
But meaningful human oversight means:
- Active involvement — humans can override the AI.
- Informed oversight — they understand the model’s logic.
- Timely intervention — review happens before any decision takes effect.
If those conditions aren’t met, adding a human later doesn’t make the process compliant — it just gives the illusion of control.
4. What lawful use requires
If Article 22 applies, organizations can proceed only when:
- it’s necessary for a contract (rare in recruitment), or
- the candidate gives explicit consent, or
- a meaningful human review step is introduced.
They must also:
- inform candidates transparently about AI use,
- enable contestation and human intervention, and
- conduct a DPIA (impact assessment) since AI screening is high-risk processing.
5. Why this matters beyond GDPR
Under the EU AI Act, recruitment and CV-screening systems are now explicitly listed as “high-risk AI systems.” This means they require:
- documented data-governance,
- human-oversight mechanisms,
- risk management, and
- conformity assessments before deployment.
GDPR still governs transparency, fairness, and data rights — but AI governance will add a new layer of accountability.
6. What this teaches us about AI literacy in HR
This isn’t just about compliance checklists. It’s about understanding how AI really works — and where human responsibility begins.
AI literacy in HR means:
- knowing what kind of decision the AI makes,
- asking who is accountable if it’s wrong,
- ensuring data quality and bias testing, and
- being able to explain the logic to candidates and managers alike.
True literacy isn’t coding. It’s the ability to challenge an algorithm with the right questions — before it quietly shapes people’s careers.
Is your screening AI making automated decisions? Find out before a regulator asks.