We (Matt Stempeck, Eloise Gabadou, and Yung-Hsuan Wu) were elated to be at The Impacts of Civic Tech Conference 2025 in Mechelen, Belgium last month.

And now that the videos are up, we can share our presentation with you:

Community-led AI audits: why they matter and how to start one

Get the slides

Audits of AI systems looked at predictive algorithms in the criminal justice system; discrimination by ride-hailing platforms; and city-level social services — and had real impact. You too can lead audits of AI systems affecting your own communities. Together we can hold these systems accountable for equitable outcomes in the real world.

We present an EU-funded partnership to audit AI systems that affect the public interest, working with Spanish company Eticas.ai to measure performance of protective attributes within data-driven decision-making systems.

By testing AI models’ actual outcomes and impacts for bias, the project can spotlight discriminatory systems. You’ll hear about audits of real-world AI systems, which looked at predictive algorithms in the criminal justice system, discrimination by ride-hailing platforms, and city-level social services — with real impact.

Also discover open benchmarking data and tools that can help you lead audits of AI systems affecting your own communities.

Together we can hold AI systems accountable and prove, as a community, whether their models produce equitable outcomes in the real world.

The Q&A for this session can be seen on mySociety’s YouTube. Thanks to Eric Reese for moderating us, and mySociety for convening us.