Resources: Response, retrieval & resistance

Auditing how AI models handle politically contested information, GlobalFact 2026

We've found that AI models differ meaningfully in whether they mirror or resist state-aligned framing.

In this workshop at GlobalFact 2026 in Vilnius, participants ran live tests across multiple models and leave with a reproducible audit methodology.

Read our research in China, and Iran (with ASL19/Factnameh).

Interested in our nifty tool, AIdas? Feedback, ideas?
Get in touch: patrick@gazzetta.xyz / valentina@conexionsegura.org

[Newsletters]