Insights · AI Adoption, Community
From ethical AI to sustainable AI.
8 October 2026 · Emma Ockelford
At an LGAi session in September, colleagues from more than 20 councils talked about the environmental cost of AI. Most councils have climate commitments. Nobody in the room was yet counting AI in them.
Most councils have spent the past two years working out how to use AI responsibly. Data protection, bias, transparency, professional judgement. Important questions, and the sector has made real progress on them.
A quieter question is now starting to surface. What does all this AI cost the planet?
That was the focus of an LGAi session in September, and it drew colleagues from more than 20 councils. Climate officers, digital leads, practitioners and people who simply wanted to understand the issue better.
One finding stood out. Nobody in the room was yet including AI use in their council's greenhouse gas reporting. Not because they did not care, but because nobody has a clear way of measuring it.
The discussion was honest about the tension. AI transcription and drafting tools are giving social workers and caseworkers real time back. That time goes into visits, relationships and better decisions. Nobody wanted to argue for taking that away.
But there was equally strong agreement that "use it because it is there" is not a strategy. Several people made the same point in different ways: right tool, right job.
A few practical ideas came out of the conversation.
Start with a baseline. At least one council is already working to understand the emissions from its technology estate. You cannot manage what you cannot see.
Ask suppliers better questions. Where is the model hosted? How is that data centre powered? Is a very large model really needed for this task, or would a smaller one do the job as well?
Agree what good use looks like. One suggestion was a shared fair-usage approach across councils, so each authority is not writing its own from scratch.
And remember the wider picture. As one participant put it, reducing local emissions means little if the impact simply moves to a data centre somewhere else.
This matches what we are learning in our own product work. When we tested different ways of producing AI-assisted drafts, the most elaborate approach was both more expensive and slightly less accurate than a simpler, more structured one. Bigger is not always better. Often it is just bigger.
We think this is where the next phase of responsible AI in local government is heading. From ethical AI to sustainable AI. The same discipline that asks "is this fair and safe?" should also ask "is this proportionate?"
None of this needs to slow councils down. It asks them to be deliberate.
LGAi is free, member-led and vendor-neutral. The conversation on sustainable AI continues at LGAi.uk. What is your council doing to understand the footprint of its AI use?