What happens when radicalization moves from public social media to private conversations with AI? Paul Ash of the Christchurch Call Foundation says it means fewer of the checks and balances found on other platforms.

In this conversation from The Paris Peace Forum, Tony Maciulis asks Ash how AI plays a role in recruitment by violent extremists. Ash says researchers long thought a conversation with an infinitely patient AI recruiter would differ from one with a human. He says it could be scaled very quickly, especially with models that have had their guardrails stripped off. He calls that probably the biggest risk area right now.
Ash says the new benchmark won't fix that problem, because those models are moving so quickly. He also says much of the real harm happens in longer conversations. That is why the benchmark uses a multi-turn approach.

This conversation is part of Global Stage, an award-winning series produced by GZERO Media in partnership with Microsoft exploring the intersection of technology, politics, and society.