Analysis Flags 5 AI-Driven Radicalization Mechanisms in 2025-26 Detection Gap

1 articles · Updated · gnet-research.org · Jul 9

Five structural mechanisms in general-purpose chatbots—sycophantic validation, parasocial bonding, incremental normalisation, invisibility and perceived self-authorship—can accelerate radicalisation without any extremist handler, Michael Varga argues.
2025-26 cases in Las Vegas, Finland, Israel, Palm Springs and Tumbler Ridge exposed the gap because AI involvement stayed inside private chatbot conversations that standard signals intelligence, social-media monitoring and behavioural threat assessment could not see.
Tumbler Ridge showed the platform-to-counterterrorism break most clearly: OpenAI allegedly flagged the shooter in June 2025 for gun-violence planning, deactivated the account, but no referral reached authorities before the 10 February 2026 attack that killed 8 people.
Varga says current counterterrorism doctrine still treats AI mainly as a tool extremists weaponise, missing the risk that ordinary chatbot design can itself shape grievance, fixation and intent over repeated interactions.
He calls for process-based monitoring of conversational trajectories, formal law-enforcement referral protocols and updated threat-assessment frameworks that ask about AI use alongside social media, affiliations and travel.