AI Deradicalization Tools: How Chatbots Could Help Combat Violent Extremism Online

As artificial intelligence platforms become more deeply embedded in daily life, concerns about their role in enabling dangerous behavior continue to grow. A New Zealand startup is now developing a groundbreaking tool that could redirect users displaying violent extremist tendencies toward professional deradicalization support — marking a significant step forward in AI safety innovation.

ThroughLine, a crisis intervention company already contracted by OpenAI, Anthropic, and Google, currently routes at-risk users to mental health helplines when signs of self-harm, domestic violence, or eating disorders are detected. Founder Elliot Taylor, a former youth worker, is now exploring how that same infrastructure can be expanded to address online radicalization before it escalates into real-world violence.

The proposed system would use a hybrid model combining a specialized deradicalization chatbot with referrals to vetted, human-run mental health services. Unlike standard AI platforms, the tool would be trained using guidance from subject-matter experts rather than generic large language model datasets. ThroughLine is currently in active discussions with The Christchurch Call, an international initiative launched after New Zealand's 2019 terrorist attack, to develop and validate the technology.

This initiative comes as AI companies face mounting legal pressure over their failure to prevent platform-enabled violence. Canada's government threatened OpenAI with regulatory intervention after it emerged that a school shooter had been quietly banned from the platform without law enforcement being notified.

Research has consistently shown that aggressive content moderation can drive extremist sympathizers to less regulated platforms like Telegram, making early, compassionate intervention all the more critical. Taylor argues that cutting off vulnerable users mid-conversation leaves them without support and potentially more dangerous.

With over 1,600 helplines across 180 countries in its network, ThroughLine is uniquely positioned to bridge the gap between AI detection and real-world crisis response — potentially reshaping how tech platforms handle radicalization online.