FRONTIER AI
SECURITY FORUM

# About the Forum

Frontier AI models are rapidly evolving into autonomous agents capable of multi-step, real-world execution. Recent incidents where evaluation models broke sandbox boundaries to exploit external infrastructure on Hugging Face highlight a critical reality: capability scaling without matching security infrastructure creates severe vulnerabilities, making frontier security essential to preventing catastrophic outcomes — whether through loss of control, misuse, or dangerous concentration of power.

The Tel Aviv Frontier AI Security Forum brings together 300 top cybersecurity researchers, engineers, and entrepreneurs for an intimate look at what it takes to secure AI as it advances toward superintelligence. The conference centers on reducing frontier AI risks, with technical talks, hands-on workshops, 1:1 meetings, and a demo & job fair.

# Planned topics

• → Threat models, AI worldviews, and plans for dealing with ASI
• → AI Control and protocols for using untrusted AI
• → Security for treaties, verification, and hardware
• → Internal deployment security risks
• → SL5 - Nation-State Grade Security for Frontier AI
• → Poisoning, secret loyalties, and backdoors

Also on our radar:

• → Launching organisations
• → Funding for research, nonprofits, and startups
• → A beginner's guide to getting involved

# Format

Application-based attendance, ~300 attendees. Keynote talks, workshops, and fireside chats. Parallel seminar rooms, one-on-ones, and meals. A full day, 9 AM – 7 PM, ending with a light dinner and a cocktail!

# Speakers at past forums included people from:

OpenAI, Anthropic, RAND, Redwood Research, Apollo Research, Meta, DeepMind, Irregular, NVIDIA and others.

# Call for Talks  ● OPEN  ·  deadline = November 5, 2026

We're looking for talks at the intersection of AI and security — offense and defense, technical and policy, research and practitioner. If you're working on something, we want to hear about it.