On the research front, new security techniques for AI systems are being explored to defend against adversarial attacks and manipulation.