Before a product, a doctrine. Cognitive security rests on a scientific foundation developed with academics and intelligence specialists: a framework to read and govern how AI enters the decision-making process.
Current regulatory frameworks — from the AI Act to corporate ethics — act on outputs. Cognitive AI Governance intervenes one level earlier: where AI structures the context in which a decision takes shape.
The algorithmic infrastructure that filters and pre-selects information before it reaches the conscious mind — beneath Kahneman’s Systems 1 and 2.
The gradual, hardly reversible outsourcing of thought to external, opaque AI systems, with loss of independent deliberation.
Four decision postures — exploratory, generative, critical, operational — to read how AI intervenes at each stage of a decision.
The index that measures an organisation’s cognitive posture across four dimensions: aggregate, organisation-level, trackable over time.
Executive experience with European boards and leaders: months of field work, over a hundred executives involved, a platform with posture switching and cognitive audit trails.
System-scale application within the CNR–AISES–SPES perimeter: training on cognitive sovereignty, technological monitoring and strategic advisory.
The Cognitive Sovereignty Index is developed jointly by the two organisations: it is the bridge linking advisory research to system-scale application.
The AI Act prohibits manipulation only beyond the threshold of « significant harm » (Article 5) and safeguards fundamental rights downstream of the output (FRIA, Article 27). What remains uncovered is ordinary, sub-threshold use — where cognitive risk accumulates day after day. The Cognitive Impact Assessment (CogIA) is the proposal to fill that space: an impact assessment complementary to the FRIA that measures, upstream of the output, how a system structures the problem, how much room it leaves for dissent and verification, and what degree of dependence it induces over time.
The fastest route: the CogIA is born as a shared code of conduct and matures into industry practice, up to the rank of recognised standard.
In the technical fora of European standardisation, the cognitive impact assessment methodology enters the toolkit of conformity.
Since February 2025, providers and deployers must ensure the AI literacy of their personnel: no standard today safeguards the cognitive dimension of that obligation.
The deferral of high-risk obligations to December 2027 widens the window: whoever defines the standard of measurement sets the terms by which the entire field will be judged.
