AI governance & disinformation
PDIP
I built a working prototype that turns suspected AI-generated or coordinated political content into a documented compliance review.
The problem
Teams reviewing suspected disinformation need to move quickly without turning an AI score into a verdict. They also need to show which rule applies, what evidence was used, and who remains accountable.
My part in the work
I designed and built the Parliamentary Disinformation Intelligence Platform as a policy and interface prototype following my work at the CIVICA EU AI Policy Hackathon.
What was involved
What I did
- I designed an intake flow for suspected synthetic or coordinated text, transcripts, and media links.
- I connected each review to relevant provisions of the EU AI Act, Digital Services Act, GDPR, and EU political-advertising rules.
- I made source evidence, escalation, human review, and an audit trail visible in the interface.
- I added red-team scenarios so the review flow can be tested without presenting generated outputs as real incidents.
What changed
- A public, working prototype that demonstrates the complete review journey.
- A plain link between model output, source evidence, regulation, and human accountability.
- Reusable test scenarios covering elections, manipulated media, and public-interest claims.
Why I include this work
I include PDIP because it shows the governance work itself—not just a certification. It makes the rules, uncertainty, and human responsibility visible in a product people can try.
See the live project