The classic trap
Recital 6 establishes the principle of human-centric AI, anchored in Article 2 TEU and the Charter. In practice, this recital serves as the interpretive lens used by the EU AI Office or the CNPD (on the personal data side) when qualifying a use case as abusive, disproportionate or harmful to fundamental rights. Organisations deploying AI systems without documenting their alignment with these values risk requalification: a borderline use case (HR scoring, customer profiling, predictive model) can shift into the high-risk or prohibited category precisely because the impact assessment never materialised the 'human at the centre' test.
The 'human-centric' test: how to make it operational
A recital cannot be ticked off, it must be argued. To turn this legislative intent into opposable evidence, your AI compliance file must concretely demonstrate:
- That the purpose of the system increases the well-being of end users, not just internal productivity or margin.
- That the Charter rights potentially impacted (dignity, non-discrimination, privacy, effective remedy) have been identified one by one before deployment.
- That the human retains real control: oversight, override capability, right to explanation, non-AI alternative for the user.
- That mitigation measures are proportionate to the risk for the person, not to the commercial risk for the company.
- That internal governance (AI committee, DPO, CISO, business owner) traces arbitrations when model performance conflicts with fundamental rights.
Without this traceability, recital 6 becomes an argument against you the day a user, a union or an authority challenges the use.
How Luxgap automates this risk
Our Luxgap Human-Centric AI Compass turns the abstract test of recital 6 into an opposable alignment dossier, automatically generated for each AI system in your portfolio. A specialised LLM agent queries your technical sources (Azure ML, AWS SageMaker, Vertex AI, HuggingFace endpoints, Copilot M365, Odoo instances with AI modules) and cross-references detected uses with a fundamental rights matrix derived from the Charter and EU AI Office guidelines.
- Continuously scans the AI systems actually deployed in your IS and identifies those that have never been subject to an alignment analysis with Union values.
- Generates for each system a fundamental rights impact sheet (dignity, non-discrimination, privacy, freedom of expression, effective remedy), pre-filled from the declared purpose and data types processed.
- Automatically detects blind spots in human control: no override, no non-AI alternative, no challenge channel, purely formal oversight.
- Classifies each system against the AI Act grid (prohibited, high-risk, limited risk, minimal risk) and instantly alerts when a borderline use emerges in a business team without prior validation.
- Produces a timestamped, cryptographically sealed PDF report demonstrating human-centric alignment, opposable to the EU AI Office and the CNPD during an inspection.
Available as a complement to a Luxgap DPO or CISO mandate or as a dedicated SaaS module depending on your scope. Request a tailored quote and our teams will prepare a demonstration on your real AI systems, with a free 48-hour blank audit to measure your exposure before any engagement.
