At the CPDP 2025 Conference in Brussels, EUCAP and AIRA was represented by Leo Vanderdonckt, who attended key sessions and workshops on how the EU s Artificial Intelligence Act (AI Act) will be implemented in practice.
His key takeaway: those most affected by AI systems including autistic and neurodivergent people remain largely excluded from policymaking, design processes, and even ethical oversight. There was almost no involvement of any minority groups in the discussions, Leo observed, after speaking with researchers and attending sessions on compliance and governance.
These insights reflect AIRA’s wider concern: while regulation is advancing, questions of participation, lived experience, and social inclusion remain unresolved. How do we ensure that AI systems respect human rights not only in theory, but in lived reality?
The AI act: Regulation in need of representation
The EU Artificial Intelligence Act is Europe’s first comprehensive law regulating artificial intelligence. It introduces legal obligations for so-called “high-risk” AI systems used in areas like healthcare, education, law enforcement, and social welfare.
At CPDP 2025, experts from across Europe came together to discuss what compliance, oversight, and enforcement might look like in practice. But Leo noted a lack of representation from groups most impacted by these systems:
Almost all the speakers or topics were related to either law or computer science. The social sciences were completely absent.
– Leo Vanderdonckt
Impact Assessment Tools: Too Abstract, Too Isolated
Leo joined a workshop on Fundamental Rights Impact Assessments (FRIAs), tools meant to help organisations assess the human rights risks of their AI systems. One scenario involved an insurance company using AI to decide eligibility and pricing for life insurance.
“I made a statement of possible discrimination against autistic people,” Leo recalled.
Although the tool showed promise, participants found it difficult to use especially when estimating risk levels numerically or anticipating real-world impacts.
The tool was based on a framework developed by Gianclaudio Malgieri and colleagues (see article) and is still evolving. AIRA welcomes this work and highlights the importance of real-world testing and inclusive input.
Participation in AI Governance: Still More Theory Than Practice
Another workshop explored how AI governance could be made participatory. Participants were asked to map where civil society involvement should occur from policy development to algorithm audits. Leo noted that many participants struggled:
A lot of us were not aware of the current situation in which there is hardly any participation from interest groups to reflect on where or how we would like participation organised.
The session, led by the LSTS research group at Vrije Universiteit Brussel, presented a promising model, but the experience made clear: participation needs structure, not just intention.
What the AIRA Survey Tells Us
Leo’s field observations resonate strongly with AIRA’s own research. Through the AIRA AI and Rights for Autistic People project, EUCAP conducted a Europe-wide survey in 2023 asking autistic people how AI and automated systems affect their lives.
Key findings from the AIRA 2023 Survey:
- AI systems frequently misinterpret autistic behaviour, especially in communication
- Emotion recognition tools were seen as particularly biased and harmful
- Respondents felt excluded from AI policy, design, and governance
AI often assumes one way of thinking, feeling, and expressing and that doesn’t work for us.
-AIRA survey respondent
At CPDP, Leo shared these findings with researchers, who expressed interest in EUCAP’s work and the AIRA survey.
What Needs to Change
EUCAP believes that AI regulation must not only define risk it must centre the people whose rights are at stake.
To achieve this, we need:
- Impact assessment tools that are clear, usable, and co-designed with those affected
- Neurodivergent voices included in all phases of AI design and oversight
- Cross-disciplinary collaboration beyond legal and technical silos including lived experience and social sciences
“If AI is to serve people, it must include them not just as users,
-Leo Vanderdonckt, CPDP 2025-
but as co-creators.”
Further Reading
- Survey and literature review 2023 -Artificial Intelligence and the Rights of Autistic People (AIRA)The EU Artificial Intelligence Act
- EUCAP’s AIRA Project Overview
- Malgieri et al. on FRIA Frameworks
- FRAIA – Fundamental Rights and Algorithms Impact Assessment (Netherlands)
- A public-sector tool used by Dutch authorities to assess how algorithms may affect fundamental rights. Co-developed with Iris Muis.
- CPDP Conference Official Site
