Preventive Safety: OpenAI Implements Extremism Filter by ThroughLine

Compliance is becoming the foundation for foundational AI models. On April 2, 2026, OpenAI confirmed the integration of solutions from New Zealand's ThroughLine to route users showing signs of violent extremism.

Instead of a simple shadow ban, the system will redirect problematic queries to specialized human or bot support. This move (backed by The Christchurch Call coalition) is a direct response to the growing demands of government regulators (recall the California filter law of March 31). OpenAI is building a defense-in-depth security system (AI Safety) to guarantee its Enterprise clients an absence of reputational risks when deploying agents in the public sphere.

Source: Reuters / Bangkok Post