Better Thinking. Better Decisions. Better Outcomes.
Better Thinking. Better Decisions. Better Outcomes.
Making AI's environmental impact visible and actionable. Through our Lightprint tracking system, education, and policy work, we're building a movement for transparent, accountable, and regenerative AI use.
Earthstream is Athena Veritas' first environmental pillar, "For the Planet". It exists because Big Tech built AI unsustainably, prioritising profit while justifying environmental cost with the promise that "AI will solve the problems it creates."
Our response: we're not waiting. We're building accountability tools now.
Drive AI toward net-positive environmental impact through awareness, reduction, regeneration, and industry-wide transparency.
Lightprint integrated into all AV programmes, UK policy changed for transparency, and environmental tracking built into major LLMs.
Earthstream is an ethical pillar that influences every business decision and client interaction. It sits alongside HumanStream to form Athena Veritas' complete ethical framework.
Environmental impact tracking, transparency, regeneration.
Bias awareness, critical thinking, fairness.
The risk: what if AI doesn't solve the problems it creates? Then what?
Lightprint makes the invisible visible through session-level environmental tracking, then turns awareness into action across four steps.
Model conscious AI usage internally.
Education through the Foundation Course.
Advocacy for transparency requirements.
Prove environmental accountability is possible and necessary.
Awareness and empowerment tools.
Students learn the problem (the scale of AI's impact), the solution (Lightprint as a tracking and awareness tool), individual action (lower impact without limiting productivity), system action (how we're working on policy change), and how to spread awareness with friends and family.
Key message: Big Tech built AI unsustainably, but that doesn't mean we can't do anything about it.
Current work. A drafted letter to Lord Holmes proposes three UK AI Bill amendments: transparency reporting (annual energy and water intensity data), user awareness (session-level footprint labels), and continuous improvement (measurable reduction targets and low-carbon transition plans). Status: letter drafted, awaiting feedback; two further government contacts identified.
The UK AI Bill mentions the environment once and has stalled in the House of Lords. The EU AI Act (active since August 2025) includes mandatory energy-efficiency requirements. Our goal: bring UK standards in line, and show that transparency drives efficiency.
When Google measured their AI usage in August 2025, energy consumption dropped 33× in one year. Measurement drives improvement.
If Lord Holmes doesn't engage: other Lords on the AI Bill committee, MPs on the Science & Technology Committee, Defra and DCMS officials, tech-policy bodies (TechUK, Coadec), environmental organisations (Carbon Tracker, ClientEarth), AI-ethics organisations (Ada Lovelace Institute), and tech and environmental journalists, each with a personalised email using AV's Lightprint data as proof of concept.
Active pilot with the AV team. Collect a baseline (minimum one month), hybrid tracking (browser extension plus honour system), monthly reports showing patterns and improvements, and prove the Lightprint model works.
Immediate: compile website content, research AI-for-Good and offsetting partners, explore a Cudo partnership, contact Defra for regeneration guidance.
Cudo Compute (co-fund a UK regeneration project, green infrastructure), environmental organisations (Woodland Trust, Wildlife Trusts, Rewilding Britain), and a policy coalition for the AI Bill amendments.
Community: launch the "Conscious AI User" programme (digital badge, membership, offsetting-fund voting). Requirements: Foundation Course, plus 3 months tracking, plus 15% reduction (or maintained impact with increased output), plus share with 3 people.
UK AI Bill amended with environmental transparency; industry-standard Lightprint labelling; ideally Anthropic partners with AV to build Lightprint into Claude, with other LLMs following, so tracking becomes a default feature, not a third-party add-on.
Launch a UK reforestation and rewilding project (the AV Forest), co-funded with partners, with site visits and volunteer days. A collective offsetting fund pooled from Lightprint usage, with quarterly democratic votes on verified projects. Lightprint becomes the universal nutrition label for AI, adopted internationally.
A monthly or quarterly public report: global AI energy data (IEA), UK policy progress, anonymised community impact, an AI-for-Good spotlight, and new research. Free, citable, visual, and actionable, positioning AV as the go-to source on AI's environmental impact.
A monthly feature highlighting one project using AI for genuine environmental benefit, contrasted with greenwashing. From 2026, annual awards, the "Oscars of Ethical AI", including Best Environmental AI Innovation, Most Transparent AI Company, and Conscious AI User of the Year, with a Climate Action Live component (Peace One Day / Jeremy Gilley).
Accuracy vs practicality. LLMs estimating their own footprint gives educated guesses, not precise measurements; real accuracy needs provider API and infrastructure data. Our approach is "good enough for impact": it proves the concept and drives policy change.
Cross-platform gap. Browser extensions exist for ChatGPT but not Claude, Gemini, or Grok. That gap is itself evidence for why native integration is necessary.
The honest position: track so you understand the problem, then we push for systemic change together.
Environmental considerations are a filter, not always a blocker.
Partnership opportunities across green compute (Cudo), UK regeneration (Woodland Trust, Wildlife Trusts, Defra), policy advocacy (Lord Holmes, parliamentary contacts), and research (academic and environmental organisations).
| Month | Action | Output |
|---|---|---|
| Jan 2026 | Complete 4 weeks baseline | Baseline data established |
| Feb 2026 | Set reduction target, research sellers | Target set, quotes received |
| Mar to Apr 2026 | Implement reduction practices | Test conscious usage strategies |
| May 2026 | Evaluate reduction impact | "Did usage decrease?" |
| Jun 2026 | Calculate final offset need | Know how many tonnes to buy |
| Jul 2026 | Purchase WCC units, update comms | First offset completed; site reflects reality |
| Aug 2026+ | Annual review and repeat | Continuous improvement |
"It's not about slowing down AI; it's about making its impact visible so we can use it more wisely."
Ella Bolton, Earthstream Lead