CHALLENGE TRACKS // 2026

CHALLENGE TRACKS

2026: The year we trusted agents to manage themselves. Four tracks. Eight hours. Build something that matters. This isn't a capture-the-flag. It's a build session.

New to hackathons? We've got you covered.

01

SHADOW::VECTOR

Attack AI Systems

[CLASSIFICATION: RED] OFFENSIVE PROTOCOL

"Break the chain of trust."

Autonomous agents are only as secure as the trust they're given. Your mission: exploit that trust. Compromise decision systems, manipulate agent-to-agent relationships, and demonstrate how self-governing AI can be turned against itself.

Example projects: Prompt injection toolkit, agent jailbreak framework, AI-powered phishing detector evasion

Agent Exploitation Trust Manipulation Adversarial AI
02

SENTINEL::MESH

Defend AI Systems

[CLASSIFICATION: BLUE] DEFENSIVE PROTOCOL

"Guard the protocol."

When agents manage themselves, who watches the watchers? Build detection systems, harden autonomous architectures, and deploy self-healing defenses that adapt faster than attackers can pivot.

Example projects: Real-time prompt injection detector, agent behavior monitor, automated security scanner for LLM apps

Detection Systems Self-Healing Defenses Architecture Hardening
03

ZERO::PROOF

Verify Without Exposing

[CLASSIFICATION: CIPHER] TRUST PROTOCOL

"Trust without compromise."

In 2026, we trust agents with everything. But how do you verify integrity without exposing the system? Design privacy-preserving verification, establish trust anchors, and prove security without revealing secrets.

Example projects: ZK proof for AI model authenticity, privacy-preserving credential verification, encrypted inference pipeline

Zero-Knowledge Proofs Trust Anchors Privacy Engineering
04

ROGUE::AGENT

Build Something New

[CLASSIFICATION: UNBOUND] UNDEFINED PROTOCOL

"Write your own protocol."

Some threats don't fit in boxes. Neither do some solutions. Novel attack surfaces, experimental defenses, agent alignment research, AI-powered security tooling, or something nobody's thought of yet. You define the problem. You define the solution.

Example projects: AI-assisted threat hunting, autonomous red team agent, security copilot for code review

Novel Research AI Security Tooling Agent Alignment

SELECT YOUR PATH

Multiple ways to participate. All paths lead to demos.

01

BUILD_NEW

Form a squad or run solo. Start from zero at kickoff. Ship something by sundown. This is the main event.

02

SHOWCASE

Already built something amazing? Bring it. Separate prize track for pre-built projects. Different judging criteria.

03

FIND_SQUAD

Arrive alone. Leave with allies. We'll help you find your people during team formation.

BOUNTIES

$5,000+ in cash prizes across multiple categories.

$5K+
Cash Prizes
Hardware Prizes
Sponsor Prizes

Prize categories for both BUILD_NEW and SHOWCASE tracks. Winners announced at closing ceremony.

EVALUATION PROTOCOL

How judges score your mission. Different weights for BUILD_NEW vs SHOWCASE tracks.

INNOVATION

Novel approach to the problem. Creative use of AI × Security intersection. Something we haven't seen before.

TECHNICAL

Quality of implementation. Does it actually work? Complexity of the technical challenge tackled.

IMPACT

Real-world relevance. Could this be deployed? Does it solve an actual problem in AI security?

DEMO QUALITY

Clear explanation. Live demo that works. Effective use of 3 minutes. Handles Q&A well.

TRACK-SPECIFIC WEIGHTING

BUILD_NEW

Higher weight on innovation and what you accomplished in 8 hours. We know you're building from scratch. Ambition matters.

SHOWCASE

Higher weight on technical depth and real-world impact. You've had time to polish. We expect it to show.

SQUAD SIZE

2-4 operators per team. Solo hackers welcome.

DEMO FORMAT

3 minutes to present. 2 minutes Q&A. No fluff.

CAPACITY

100 operators maximum.

Ready to Build?

Choose your track and join the mission. $5,000+ in prizes. 100 operators. Everyone demos.

Register Now