AI models have had a ~4x YoY increase in compute for the last 70 years. In the security domain, what has 4x effective compute brought us in 2025 and what will it bring us in 2026? In this session, Jason will give us a survey of the bleeding edge of security applications, from a frontier AI lab perspective, including advanced persistent threats, and what new security threats are coming from AI and can be defended by AI in 2026 and beyond.