Environmental policy

Last updated June 2026.

ASoc recognises that AI model training and inference carry a real energy cost. We are committed to minimising that footprint and being transparent about what we do to reduce it.

Infrastructure

We prioritise cloud providers that source a significant share of their electricity from renewable energy, and we right-size compute to the task rather than over-provisioning by default.

Model efficiency

We prefer the smallest model that meets the accuracy and latency requirements of a project. Compact, well-tuned models use less energy per inference than larger general-purpose alternatives.

Continuous improvement

We review our tooling and hosting choices at least annually and report internally on estimated carbon outputs, with a goal of reducing them year-on-year.

Questions or suggestions? Email us at hello@asoc.co.