Use computing thoughtfully
Choose models and infrastructure to fit the task. Consider energy, storage and repeated processing alongside usefulness.
Useful intelligence should help people while taking the resources it needs seriously. We want that responsibility to shape our choices from the beginning.

These intentions will guide our work as it develops.
Choose models and infrastructure to fit the task. Consider energy, storage and repeated processing alongside usefulness.
Use remote collaboration and digital documents where they serve people well and avoid unnecessary travel or material use.
Consider repair, reuse and refurbishment before replacement, with responsible recycling when equipment reaches the end of its life.
Ask for clear evidence about environmental practices, sourcing and resource use when evaluating technology partners.
Use evidence to understand impact. Share what has been measured, the assumptions involved and what remains uncertain.
Consider the wider consequences of a technology choice, including access, community needs and the environment.
These are the choices we intend to make, rather than a report of measured environmental performance. AI has a resource footprint; its value needs to be considered alongside that cost.
We will make claims about reductions, carbon neutrality or net zero only when there is evidence to support them.