World Economic Forum
A framework for responsible AI adoption, scale and management in social innovation.
A design guide when you plan & build AI systems — and a diagnostic map when an AI deployment goes wrong and you need to find out where.
Built on existing research, policy and practice in Responsible AI, and shaped for organisations whose mission is social or environmental — not scale for its own sake.
THE TOOL IMPACT TEAMS NEED IN THE AGE OF AI
The Responsible AI Compass is a simple map of the four areas where conscious decisions have to be made when a team uses or builds AI: your purpose, the tech behind it, the governance around it, and its effect on people & planet. It gives you the questions worth asking, before launch and again when something goes wrong.
It exists because there is still no agreed standard for doing AI responsibly, and most harm shows up quietly: a group left out of the data, a team disengaging, a decision nobody can explain, a cost nobody counted. Teams that ask these questions regularly build the kind of trust that beneficiaries, donors, partners, and governments respond to. Responsible AI isn't just risk mitigation — it's how you invest in your long-term impact.
The goal is not perfection. It’s progress, made visible.
BEFORE DEPLOYMENT
Before you build: name who this will affect, including people who won't be in the decision room. Say who has access and who is accountable when something fails. Check whether the technology does what you intent it to do. Then ask whether any of this actually serves the communities you exist for. If you can't answer all four, you're not ready.
AFTER DEPLOYMENT
Don't start with the error log. Start with who was harmed/ or is at risk and then trace it back: what decision allowed this? What governance gap meant no one caught it? What design choice created the condition? You'll usually find the technology did exactly what it was built to do - the problem was built in earlier.
The compass
Before anything else: what is the organisation for, and why does this problem need AI at all? Purpose is the conversation most teams skip — and the one every failure eventually traces back to.
When it goes wrong here
When something goes wrong here, the system usually worked as built - it was simply built for the wrong reason, or with weak understanding of the why..
World Economic Forum
A framework for responsible AI adoption, scale and management in social innovation.
NIST
A framework for managing AI risks and fostering trustworthy AI development and use.
OECD
Values-based principles and policy recommendations for trustworthy AI.
European Union
The official regulation establishing a risk-based legal framework for AI.
Partnership on AI
Research and practical guidance for the responsible development and use of AI.
CDAC Network, Alan Turing Institute & Humanitarian AI Advisory
A governance and assurance framework for humanitarian organisations using AI.
GCRAI
Resources on AI governance, responsible implementation and ethics.
Your turn
If a question saved you from a bad decision, or a layer failed to catch something real, that belongs here.