The dominant frame for AI governance — the one inherited from US security
compliance and bolted onto European regulation — measures three things:
whether the system is lawful, whether it
respects individual data rights, and whether
the organization can document what it built.
These are necessary. They are not sufficient.
A trained model is also a thermodynamic event. The training of a single
frontier model can emit hundreds of tonnes of CO2e. Inference
at deployment scale routinely dominates the lifetime energy budget of the
system. Data centers in the Netherlands alone consume measurable
fractions of the national grid. None of this is captured by the
three-axis view.
BesaGreenTech adds the fourth axis. We assess AI systems against
EU AI Act conformity, ISO/IEC 42001
management-system maturity, and GDPR data-protection
obligations — and, in the same engagement, against energy intensity,
embodied carbon, lifecycle hotspots, and the emerging
CSRD reporting expectations on digital infrastructure.
One report. One evidence chain. One coherent picture of what the AI
system does, what it costs the public, and what it costs the planet.