AI for Counter Extremism

Hosted by Hedayah

The rapid evolution of AI, especially generative models, presents both emerging threats and exciting opportunities in the field of countering extremism and violent extremism. Malicious actors are experimenting with ways that AI can be used as a tool to support their tactics and amplify their narratives. Similarly, AI holds potential to be leveraged as a tool to counter the threat posed by extremists by enhancing content analysis and moderation, early analysis, strategic communication, among other uses. This session will build on Hedayah’s recently published research brief - Artificial Intelligence for Counter Extremism - that offers an evidence-based assessment of AI risks, opportunities, and other policy considerations for the field of countering extremism and violent extremism. It will delve into the applications of AI for counter extremism in key emerging use cases, with presentations on these issue areas or case studies at the beginning of the workshop followed by teamwork to develop out a proposed use of generative AI for counter extremism, based on the teams’ own needs as researchers and practitioners or needs that the team have identified, with facilitated ideation, concept development, and risk assessment.

Elevating AI Safety: From Content Moderation to Benchmark Development

Hosted by the Christchurch Call Foundation

The Christchurch Call Foundation (CCF) will lead an interactive half-day workshop exploring the intersection of artificial intelligence and terrorism and violent extremism (TVE) prevention. This session will focus on two critical challenges: developing effective TVEC policies for open-source AI safety reasoning models and creating AI safety benchmarks to measure risks like grievance amplification and radicalization to violence in AI chatbot interactions. Participants will work directly with tools currently under development through CCF's Elevate project, providing valuable feedback that will be used to update and improve these tools over time.

The workshop will be structured in two segments, each featuring small group exercises. In the first segment, participants will test TVEC policies using open-source models in a sandbox environment, evaluating how AI interprets and applies content moderation policies. The second segment will focus on CCF’s effort to create an AI safety benchmark related to grievance amplification and radicalization to violence. Participants will review evaluation criteria, assess test scenarios, and annotate sample outputs from our minimum viable version of the benchmark. By the end of the session, participants will have contributed directly to shaping tools for responsible AI use in preventing the spread of terrorist and violent extremist content while gaining practical insights into emerging AI safety challenges.