AI is not free of consequence. Running a model uses real energy, water, and carbon. Anren's first commitment is to the truth, so here is that cost, stated as plainly as it can be, including what isn't known yet.
Each conversation with Anren runs a large AI model in a data center somewhere. That draws electricity, and the data center uses some water to stay cool. There is no version of this that is weightless.
Independent estimates put a single AI text reply somewhere between a fraction of a watt-hour and a few watt-hours of electricity, about the order of a short web search, plus a little water for cooling. But the figures vary enormously between models and providers, and they're debated; any single confident number is over-claiming. A reflective session is many replies, plus the background work that keeps a person's record current. So the honest framing is: small per message, real in aggregate, worth taking seriously.
No invented "your footprint is X" number will appear here until it can be stood behind. Anren is built on never fabricating, and a made-up sustainability statistic is exactly the comfortable fiction it refuses: dishonest, and how greenwashing works. The measured figure lands here the moment the data is solid, and not before.
Anren already tracks the work behind every session. That data yields an estimate of the energy and carbon each person accounts for, drawn from the published figures of the infrastructure it runs on, not from flattering math. Once the method is solid enough to stand behind, the average per-person footprint appears here, and updates as it changes.
—per person using Anren · publishing once measured
Compute is the footprint, so using less of it is the most direct lever. In practice: the smallest model that can do a given job well, not the largest by default; caching, so the same work isn't paid for twice; and no engagement patterns that keep someone here longer than is good for them. Anren's whole design already pushes against over-use. Life is what's actually lived, wherever it happens, and that principle turns out to be an environmental one too. The less Anren has to run, the better, on every axis.
Whatever footprint remains after measuring and minimizing, the commitment is to offset more than 100% of it, net-positive rather than neutral, through credible, verifiable providers rather than cheap paper credits. Better to over-correct.
The wording is deliberate: a stated commitment, not a finished accomplishment. When offsets are purchased, the provider, the amount and the proof appear on this page. Until then, no halo is claimed that hasn't been earned.
The choice of foundation is deliberate, not incidental.
Among the labs building frontier AI, Anthropic treats safety and accountable development as the point of the work rather than a constraint on it. That's the posture Anren holds toward the people in its care. Something this intimate deserves ground that takes the risks of AI seriously, and that alignment is why Claude is the model underneath. As things change, the foundation gets reassessed against the same commitments; for now, it's worth naming out loud.
It's a fair question, and it has no glib answer. AI carries real costs beyond carbon: the labor that goes into building these systems, the data they were trained on, the way the technology concentrates power. None of that gets waved away here.
A tool isn't good or bad in the abstract. It's good or bad in how it's built and what it's pointed at. Anren is pointed at something worth the cost: helping people come home to themselves and live their fullest lives, without ads, without selling anyone's data, without engineering compulsion. Better that a person spend less time with Anren and more of their life out in the world. If the technology that makes Anren possible is going to exist regardless, at least some of it should be used with this much care.
This is held with humility. Parts of it could be wrong, and the thinking will sharpen as the field, and the science on its impact, matures. What stays constant: the question keeps being asked honestly, the numbers stay transparent, and what's learned changes what's done.
Corrections, better data sources, and offset providers worth a look are welcome.
support@anren.ai