The construct is open. Attack, defend, or build something the system's never seen.
No pitches. No spectators. Just people who ship.
Not an AI hackathon. Not a security hackathon. Both: build AI that defends, break AI to expose it.
Everyone who comes builds. Everyone who builds demos.
Workshops on building, securing and breaking AI agents. Beginner to advanced.
San Francisco's AI and security builders in one room. Find your people.
Build from scratch, showcase what you've built, or come solo and find a team.
Learn from practitioners who've shipped AI security at scale.
Watch the demos — Real tools. Real exploits. Real defenses.
PRIME was the prototype (2025). SINGULARITY was the upgrade (2026). EMERGENCE is next.
New here? The cyberpunk-speak, decoded.
A build sprint: form a team (or go solo) and ship something working by the end of the day.
Software that uses AI to act on its own: browse, write code, make decisions.
Three minutes to show what you built. Working beats polished.
A theme for your project. Each has its own judging criteria.
Still have questions? Reach out. We want builders of all experience levels.
2027 tracks are announced with registration. Here's what 2026 builders took on.
Attack AI Systems
Break the chain of trust. Exploit autonomous agents, manipulate decision systems, compromise AI-to-AI relationships.
Example projects: Prompt injection toolkit, agent jailbreak framework, AI-powered phishing detector evasion
Defend AI Systems
Guard the protocol. Build detection systems, harden architectures, deploy self-healing defenses.
Example projects: Real-time prompt injection detector, agent behavior monitor, automated security scanner for LLM apps
Verify Without Exposing
Trust without compromise. Design privacy-preserving verification, establish trust anchors, prove security without revealing secrets.
Example projects: ZK proof for AI model authenticity, privacy-preserving credential verification, encrypted inference pipeline
Build Something New
Write your own protocol. Novel attack surfaces, experimental defenses, AI security tools nobody's thought of yet.
Example projects: AI-assisted threat hunting, autonomous red team agent, security copilot for code review
You own what you build. Period.
New builds start at kickoff. Open source and public APIs are fair game.
Respect the mission and each other. We're building a community, not just shipping code. Read the full protocol →
All levels and ages (minors with a guardian). AI × Security experience gets priority; passion counts too.
A working project you can show employers, investors or collaborators. Every demo is recorded.
Builders who care about AI security as much as you do.
Techniques from practitioners you won't find in tutorials.
Every team demos live. That room is the prize.
Registration opens soon. AI × Security experience gets priority; newcomers welcome.
Get Notified for 2027