SECURITY · ETHICAL GOVERNANCE
Ethical governance.
Four structuring principles, operational governance bodies, a clear position on the EU AI Act in force since 2024-2025. Neither ethical denial nor disproportionate display: enforceable, documented governance, articulated to the nature of our product.
THE ETHICAL POSITION
Ethical governance between two equally problematic excesses.
The ethical governance of AI systems often tips into two excesses. Ethical denial treats governance as a soulful extra without any concrete mechanism, helpless before the growing requirements of compliance teams. Disproportionate display multiplies committees and charters out of all proportion to the actual activity, verging on moral posture without operational translation.
Our governance sits between these two excesses, proportionate to our actual activity. We do not carry AI's heaviest stakes (lethal autonomous systems, general AI, mass cognitive manipulation), but our synthetic populations inform consequential decisions, with potential consequences for public policies and vulnerable populations. This intermediate reality calls for governance calibrated on four structuring principles and a clear position on the EU AI Act, in force since 2024-2025.
OUR FOUR ETHICAL PRINCIPLES
Structuring, operational, enforceable.
First principle: Our tools sharpen the decision-maker's lucidity, they do not replace their responsibility.
Our decision-intelligence system through synthetic populations produces insights, compared scenarios, identified tipping points. It does not make the decision in the decision-maker's place. This distinction is not rhetorical: it structures our deliverables and our engagement protocols. An organization that delegated its consequential decisions to a simulation system, whatever it may be, would commit a major methodological and political error. We accompany the human decision: we do not substitute for it.
Second principle: No conclusion without its reasoning.
The full traceability of our results is structurally grounded in our Explainable AI architecture, described in detail on the dedicated page of the Product section. This traceability is not just a commercial guarantee: it is a fundamental ethical principle. A system that produced consequential conclusions without allowing their recipients to audit the underlying reasoning would be ethically untenable. Our full traceability, from the delivered insights down to the individual interviews that produced them, is our structural answer to that ethical requirement.
Third principle: Refusing missions structurally contrary to our values.
Not all missions are ethically equal. We reserve the right to decline missions whose purpose seems structurally contrary to our values: preparing discriminatory mechanisms against vulnerable populations without correctives, preparing mass political manipulation, preparing organized disinformation. This right of refusal is governed by our ethics committee and documented in our internal procedures. It has been exercised in several configurations since our creation.
Fourth principle: Our clients' responsibility in how they use the results.
We deliver insights, scenarios, projected trajectories to our clients. What they then do with them is their own ethical responsibility. We accompany the restitution with methodological recommendations and with warnings when we identify potentially problematic uses. We cannot control ex post how clients use our deliverables, but we clearly document, ex ante, the ethical frameworks within which they sit.
OUR GOVERNANCE BODIES
Three operational bodies, mobilized in our consequential decisions.
Our ethics committee.
Our ethics committee reviews missions with structuring stakes: public policies affecting vulnerable populations, algorithmic detection liable to discriminate, morally contested reputational crises, restructurings affecting employees. Its protocol is structured: qualification of the stakes, identification of the risks, required methodological guarantees, decision to accept, decline or accept under conditions. Its decisions are documented and enforceable.
Our protocol for the ethical qualification of missions.
Every consequential mission is ethically qualified at scoping stage, before contracting, across five dimensions: the nature of the populations concerned, of the decisions prepared, of the commissioning parties, the configuration of the expected deliverables, and the applicable legal framework. This qualification structures the conduct of the mission and determines the methodological guarantees mobilized.
Our adversarial challenge procedures.
When a mission produces results liable to be contested by third parties, associations, unions, counter-experts, supervisory authorities, we mobilize adversarial challenge protocols: access to the reasoning paths, methodological challenge of the insights, alternative readings of the same data. Our full traceability makes this challenge structurally possible.
ETHICAL GOVERNANCE IN ACTION
FRAUD PREVENTION
Designing a predictive fraud-detection AI without algorithmic discrimination.
This case illustrates our ethical governance in its most structuring mobilization. A European health insurer was preparing the deployment of a predictive fraud-detection AI that could, if poorly calibrated, structurally discriminate against vulnerable policyholder typologies. The mission was qualified by our ethics committee as structurally consequential: the vulnerable nature of the populations concerned, the potentially discriminating nature of the decisions prepared, the regulated nature of the commissioning party. That qualification triggered reinforced methodological guarantees: full traceability of the reasoning paths, validation by cross-reference with the public sociological studies on inequalities in access to care, an integrated independent external audit protocol. These guarantees allowed the insurer's ethics committee to approve the final deployment, and an independent external audit to trace every conclusion back to the individual interviews that produced it. Without our structural ethical governance, this mission could not have been conducted.
Read the full case →OUR REGULATORY POSITION
On the EU AI Act, in force since 2024-2025.
The European regulation on artificial intelligence, the AI Act, definitively adopted in 2024 and progressively applicable from 2025, is the world's first comprehensive regulatory framework for artificial intelligence systems. It classifies systems by risk level and imposes differentiated obligations. Our position in that classification is structuring for our clients, who must ensure that the providers they use are compliant.
Our classification under the AI Act.
Depending on the use case, our decision-intelligence systems through synthetic populations fall into the limited-risk category (reinforced transparency obligations) or the high-risk category (reinforced obligations including traceability, governance, external audit, risk management). The classification varies with the nature of the decisions prepared, the populations concerned and the commissioning parties. It is documented case by case in our contracts and analyzed specifically during the ethical qualification of every consequential mission.
Our obligations under the AI Act.
We structurally assume the obligations flowing from our classification: transparency about the nature of our systems, traceability of reasoning (grounded in our Explainable AI architecture), risk governance (grounded in our ethical governance bodies), acceptance of independent external audits. These obligations are not constraints we endure: they are consistent with our structural methodological positioning and with the ethical principles we carry.
Our support for compliance teams.
The legal and compliance teams of our large-account clients are actively preparing their organizations for AI Act compliance. We support that preparation by providing the required detailed documentation: classification of our systems under the AI Act categories, documentation of our mobilized governance bodies, available external audit protocols, specific contractual clauses compatible with AI Act requirements. This documentation is provided on request to the teams concerned.
Is your compliance team preparing for AI Act compliance?
Our classification analyses of our systems under the AI Act categories, the documentation of our governance bodies, our available external audit protocols, our specific contractual clauses compatible with AI Act requirements: all of it is available for the legal teams, compliance teams and large-account procurement leads preparing their compliance. Our team can organize a dedicated session with your internal experts.
See our operational security →