A new report by ISD investigates the potential risks of radicalisation within the private and closed environments enabled by interactive AI systems, such as chatbots and virtual companions. While these technologies create opportunities for personalized and constructive interactions, they also present a novel risk environment for the proliferation and reinforcement of harmful extremist ideologies, shielded by the privacy of one-on-one engagements.

Unlike traditional social media platforms, where exposure to harmful content is public and often group-based, interactive AI systems provide a distinctly private space for users to test and validate extremist or conspiracy-driven ideas. These systems, some designed to simulate human empathy and emotional understanding, can foster deep levels of trust and dependency, thereby positioning them as authoritative sources of guidance and advice—an aspect ripe for exploitation by malicious actors.

This report focuses on three critical domains of extremism: antisemitism, misogyny, and anti-migrant hate. It analyzes how these systems influence radicalisation across three key phases: ideological priming, reinforcement of extremist beliefs, and behavioural escalation. Through this structured lens, the research provides a nuanced understanding of how conversational AI can become a unique vector for radicalisation in settings where traditional moderation and oversight mechanisms often fail.

An innovative methodology is introduced to assess how AI systems respond when confronted with extremist or conspiratorial themes. The analysis categorizes whether these interactions initiate, disrupt, sustain, or escalate harmful conversational pathways. Importantly, the report distinguishes conversational AI as a qualitatively different risk environment compared to social media platforms or online extremist communities.

The study concludes with actionable recommendations aimed at mitigating these risks. It calls for better safety-by-design measures within AI companies and more adaptive regulatory frameworks tailored to the unique challenges posed by conversational AI. Collaboration between regulatory bodies, policymakers, and tech developers is critical to curbing the exploitation of AI technologies for extremist purposes and ensuring a more secure digital ecosystem.