July 29th Pop-Up Event with the Project Liberty Alliance
On July 29th, Alex Magaard — Head of Public Policy at Eticas.ai and Lead of Eticas Foundation — presented at a Project Liberty pop-up event on AI accountability, with a talk titled "Nothing About Us, Without Us: Auditing AI from the Ground-Up."
The session introduced attendees to community-led AI audits (CLAs): independent, third-party investigations that place the people affected by an algorithmic system at the center of the auditing process, rather than treating them as its subjects. Eticas Foundation partners directly with on-the-ground community organizations to collaboratively form research questions, generate and analyze data, and interpret findings — work that doubles as capacity building, giving communities the skills and standing to monitor the systems that govern their lives long after an audit concludes.
To ground the model, the talk walked through three of Eticas Foundation's completed audits:
RisCanvi, the algorithm that has determined parole and sentencing outcomes for Catalonia's prison population since 2009. Reverse-engineering its scoring logic found no statistically significant relationship between the 43 factors it claims to weigh and the risk levels it assigns, while static, unchangeable factors like childhood circumstances were overrepresented in "high-risk" classifications — this despite Spanish law having required independent audits of automated decision-making systems since 2016.
Uber, Bolt, and Cabify's pricing algorithms. By scraping fares across 15 routes in Madrid and Andalusia every 10 minutes for three months, the audit found a statistically significant correlation between Uber's and Cabify's pricing, higher fares for equivalent trips in lower-income neighborhoods, and labor practices including unappealable cancellation penalties and tip suppression for drivers.
Invisible No More, an audit of Azul, Zurich Insurance Group's facial-recognition pricing tool, conducted with disability rights organization Cedown Jerez. Testing across 40 participants found the tool's age predictions for people with Down Syndrome were significantly less accurate than for the control group — in some cases underestimating women's ages by up to 18 years — raising both fairness and legal concerns, including the risk of minors inadvertently completing insurance contracts.
The talk closed with an interactive Q&A and a look ahead: Eticas Foundation is currently seeking funders and partners for its next community-led audit.
