An independent layer for observing and measuring AI behavior
NeoMundi Recherche is developing an independent infrastructure dedicated to measuring the real-world behavior of generative artificial intelligence systems.
Its mission is to document observable variations in AI outputs and their evolution over time to inform monitoring, auditing and governance. The scope of the indicators depends on the protocol, measurement version and available evaluators.
Generative systems already influence decisions, services, knowledge, and organizations. Their behavior may vary across models, providers, time periods and usage contexts. Documented observations make it possible to study these variations.
Scientific & Advisory Committee
NeoMundi Recherche brings together complementary profiles from research, education, auditing, law, and AI governance.
Their role is to help challenge methods, protocols, hypotheses, and published results, strengthening the scientific, technical, and institutional rigor of the program.
The Committee is not intended to operationally manage the Observatory. Its role is to provide an external, critical, and methodological perspective: challenging assumptions, identifying blind spots, examining reproducibility, and maintaining a strong standard of auditability.
Members
Thomas Hormaza Dow — Founder of the Business Physics AI Lab, Professor of Business With AI at Champlain College, and co-author of the REACT Framework – Cadre RÉAGI, focused on developing judgment in the use of AI in higher education.
Joël Ignasse — Science journalist for La Recherche. Author of Sur les traces du Nouveau T.rex (Eyrolles, 2025).
Cédric Chatelain — Quality consultant and IRCA-certified auditor specializing in auditing, validation, and ISO standards. Based in Bern, Switzerland.
James Aull — Founder of ASRO™, an independent witness and governed-state evidence layer for AI systems. Based in Twin Lake, Michigan, United States.
Inès Ramoul — GDPR & AI Act lawyer specializing in digital law, data protection, and AI governance. Legal and compliance contributor to the NeoMundi AI Observatory. Based in Lyon, France.
A measurement-based approach
NeoMundi does not simply observe generative systems: we measure them over time.
We develop and publish documented protocols, aggregated data and open analyses to enable critical examination and replication attempts. Campaign conditions and limits on comparability must accompany the results.
This approach follows a behavioral metrology approach for AI systems: defining indicators, documenting their calculation conditions, and studying their repeatability, sensitivity and usefulness for response analysis.
An independent infrastructure
Independence is reflected in the choice of infrastructure, the progressive openness of methods, and the involvement of external perspectives.
Infomaniak’s Swiss cloud is among the infrastructures used by the Observatory. Hosting, processing and data retention conditions are specified for each deployed setup.
This separation helps preserve NeoMundi’s ability to produce measurements independently from the systems being measured.
Public and longitudinal observations
The Observatory’s work is progressively published in several forms:
- comparative public mappings of generative AI systems;
- observation bulletins and AI Weather, according to the protocols and schedules specified in each publication;
- longitudinal analyses tracking the evolution of the same system over time;
- specialized studies by application domain;
- technical publications, datasets, and open methodologies.
NeoMundi Recherche does not seek to reduce AI systems to a ranking or a single score.
The objective is to build a common foundation for measurement, traceability, and public discussion, allowing the behavior of generative systems to be observed, compared, documented, and used as signals by governance infrastructures.