Principles for deploying AI in environmental decision-making
AI increasingly shapes how nature is monitored, valued and financed, and how countries report progress towards shared targets. CARE and FAIR set out how data should be governed. PACE does the same for the algorithmic decisions built on that data.
It has four parts: Participation, Accountability, Consent and Equity. Each applies wherever an AI system makes or shapes a decision that affects an ecosystem and the people who depend on it.
The principles are a draft. Tell us where you agree, where you do not, how you would change them and what is missing.
Drafted at the AI4PG workshop Decision Sovereignty for Nature, ETH AI Center, Zurich, 22 July 2026, and written up by Millie Chapman and David Dao.
The four principles
- Principle one
Participation
People take part in the decisions that shape their places, from setting the goals to reviewing the results.
- P1Decide close to the place
- P2Participation is ongoing
- P3Shape the goals and values
- Principle two
Accountability
Every consequential decision has someone who is held to account for it. Nature cannot file a complaint, bring a suit or cast a vote, so someone must answer on its behalf.
- A1Accountability is in place before deployment
- A2Move at the speed of trust
- A3People ratify the decision
- A4Make decisions as interpretable as possible
- Principle three
Consent
People and communities keep a standing right to say no.
- C1Free, prior and informed consent
- C2Remain unmodeled
- C3Credible exit
- Principle four
Equity
The benefits and burdens of a decision are shared fairly among the people it affects.
- E1Fair along the whole chain
- E2Share the value
- E3Credible neutrality
What do you think?
Vote on each principle and sub-principle: agree, disagree or unsure. After each vote you see how everyone else voted, and you can say how you would change it. At the end, add what you think is missing. Others then vote on it too. It takes about five minutes.
How this works
- Who can take part?
- Anyone: delegates and observers at CBD COP17 in Yerevan, and everyone else who cares how AI is used for decisions about nature. We ask where you are and in which role, and nothing else.
- What happens to my answers?
- They are published without your name and without anything that identifies you, under CC0, so anyone can reuse them. Once you finish you see all results and can download them. Each browser takes part once, and a vote cannot be changed once given.
- What happens to what I write?
- A model checks every comment and proposed statement for abuse, hate, personal details and spam before it is shown, and readers can report what gets through. Every opinion is allowed. Proposed statements join the vote for everyone who comes after you, as in Polis.
- What are opinion groups?
- Once enough people have voted, the results page sorts them into groups of people who voted alike, the way Polis does. It shows where the groups agree with each other and where they split, so the next draft can build on common ground and look again at the rest.
- Who runs this?
- AI4COP, an initiative of the Youth Negotiators Academy and GainForest. The principles were drafted at the AI4PG workshop at the ETH AI Center in Zurich and are launched for comment at CBD COP17.