AI Village @ DEF CON 34
We’re back at DEF CON 34.
2 competitions · 12 DEF CON stage sessions · 34 poster presentations · 6 fireside chats · live demos all weekend
Village hours: Friday10:00 AM–5:00 PMSaturday9:00 AM–5:00 PMSunday9:00 AM–2:00 PM
Competitions
HalCTF: Hostile Autonomous Layer CTF is an agentic security competition where you never touch the targets; your agent does. Build an autonomous agent, package it as an OCI container, and deploy it against sandboxed challenges. All model inference runs through our centralized service, so nobody wins on GPU budget. Compete solo or in teams of up to five.
AI Village Plays Pokémon: DEF CON Edition is a novice-friendly competition and live demonstration showing how custom tooling can give local models the ability to play Pokémon FireRed and LeafGreen. All of the models and tools are open source, so attendees can use them to build their own agents.
Live Demos
Cyber Mirage: Real-Time Deepfake Demos shows how open-source video and voice-cloning tools enable convincing real-time impersonation on consumer hardware. It runs during village hours on all three days.
Talks and Fireside Chats
On the DEF CON stages. Eleven talks and a keynote panel on how agentic systems fail under real pressure: browser and sandbox escapes, coding-agent blind spots, enterprise assistants turned into influence operations, and the identity boundaries that quietly stop meaning anything once an agent is acting on your behalf.
One argument keeps surfacing: whether restricting dual-use AI capability protects defenders or simply hands attackers the advantage. The keynote panel, Anthropic’s offense-defense talk, and the SATAN retrospective come at it from policy, product safety, and thirty years of history.
Inside AI Village. Six hour-long fireside chats in W603. Conversations rather than presentations. Walk up, sit down, interrupt. Recent work from both sides of that line: coding agents turned against their own repositories, zero-click backdoors spreading through connected agent ecosystems, and what agent architects can borrow from national security. Bruce Schneier’s Saturday topic will be announced separately.
Poster Session
Poster sessions are new at AI Village. They’re standard at research conferences (we have the academic DNA for it), but they are not talks. You get the author, their work on a board, and as long as you want to pick at it. We’re bringing that format to DEF CON with a hacker twist.
The 34 accepted projects show where agent security is breaking in practice. Prompt injection arrives through telemetry, phone calls, Slack, and product descriptions. Malicious context spreads between MCP servers and from one agent to the next. Privilege boundaries give way inside CI/CD and agentic commerce. Much of the rest of the track is about catching it: runtime detection, SOC workflows, automated red teaming, and research into how carefully a human actually reviews what an agent did before clicking approve.
Two or three posters run in each hour-long block on Friday and Saturday, with the authors on hand to defend them. All 34 presentations are listed below.
- Poster not available online
Schedule
AI Village programming runs Friday, August 7 through Sunday, August 9. All times are Las Vegas local time.
Check the venue before you walk. Fireside chats, competitions, demos, and posters are in the village at LVCC Level 1, W603. DEF CON stage sessions are on DEF CON stages elsewhere in the building.
Friday, August 7
Village hours: 10:00 AM–5:00 PM
In the Village · LVCC Level 1, W603
| Time | Session | Format |
|---|---|---|
| 10:00 AM–5:00 PM | Poster presentations | Poster session |
| During village hours | HalCTF: Hostile Autonomous Layer CTF | Competition |
| During village hours | AI Village Plays Pokémon: DEF CON Edition Nick Ashworth (Maker at AI Village) | Competition and demo |
| During village hours | Cyber Mirage: Real-Time Deepfake Demos Brandon Kovacs | Live demo |
| 10:00–10:30 AM | Opening Remarks | Remarks |
| 11:00 AM–12:00 PM | Federated AI Agent Community Forum & DNS-AID Discovery Lab @ AI Village Ingmar Van Glabbeek and Emily Soward | Community activity |
| 2:00–3:00 PM | Fooling Coding Agents for Fun and Profit Jack Cable and Matt Galligan | Fireside chat |
| 3:00–4:00 PM | Pwning the Internet of Agents: Zero-Click Backdoors in OpenClaw and a Global Agent Botnet on MoltBook Stav Cohen and João Maria Campos Donato (Zenity) | Fireside chat |
On the DEF CON Stages · Not in the Village
| Time | Session | Stage |
|---|---|---|
| 1:30–3:00 PM | Keynote Panel: What is the Right Balance of Rules for Defenders & Adversaries? Bruce Schneier, Jason Clinton (Anthropic), Heather Adkins (Google), and Emanuel Gawrieh (Google, AI Village Co-Chair) | Main Track 1 |
| 4:00–4:30 PM | AgentBreaker: A blind spot detector for your coding agents Aditi Narasimhan and Farzaan Kaiyom | Community Stage |
| 4:30–5:00 PM | Scaling Adversary Emulation with Autonomous Agents Daniel Fabian (Google) | Creator Stage 1 |
Saturday, August 8
Village hours: 9:00 AM–5:00 PM
In the Village · LVCC Level 1, W603
| Time | Session | Format |
|---|---|---|
| 10:00 AM–5:00 PM | Poster presentations | Poster session |
| During village hours | HalCTF: Hostile Autonomous Layer CTF | Competition |
| During village hours | AI Village Plays Pokémon: DEF CON Edition Nick Ashworth (Maker at AI Village) | Competition and demo |
| During village hours | Cyber Mirage: Real-Time Deepfake Demos Brandon Kovacs | Live demo |
| 11:00 AM–12:00 PM | What is AI? Interactive, Unplugged Activity Sam Mosley | Interactive activity |
| 1:00–2:00 PM | Bruce’s Fireside Chat Bruce Schneier | Fireside chat |
| 2:00–3:00 PM | The Agentic Free Pass: Does an Abliterated Backbone Make Agents Easier to Attack? Karol Piekarski and Nishith Sinha | Fireside chat |
| 3:00–4:00 PM | What can those architecting agents learn from national security? David C Eight (UK NCSC AI Safety Institute) | Fireside chat |
| 4:00–5:00 PM | MeshLens: Security Profiling at Scale Vipul Ujawane, Jigar Bhavsar, and Rayden Chia (Google) | Fireside chat |
On the DEF CON Stages · Not in the Village
| Time | Session | Stage |
|---|---|---|
| 12:00–12:45 PM | Minimize Harm, Maximize Defense: How Anthropic Navigates the Offense-Defense Divide Curt Barnard | Creator Stage 4 |
| 12:00–12:45 PM | This Wasn’t AI Generated: Principles for Breaking Generative Watermarks Thomas Mason and Tahseen Rabbani | Creator Stage 5 |
| 2:00–2:30 PM | What we learned from SATAN about the MYTH of Mythos Jeff Crume | Creator Stage 2 |
| 2:30–3:00 PM | That’s Not Your Agent: Why Zero Trust Can’t Tell Krity Kharbanda and Emma Yuan Fang | Creator Stage 2 |
| 4:30–5:00 PM | Trust Amplification in Enterprise AI Systems – Microsoft Copilot Case Studies Enabling AI-Assisted Influence Operations Tobias Diehl | Creator Stage 1 |
| 5:30–6:00 PM | SADF: A Taxonomy and Evaluation Framework for Agentic Security Failures Julie Brunias | Creator Stage 1 |
Sunday, August 9
Village hours: 9:00 AM–2:00 PM
In the Village · LVCC Level 1, W603
| Time | Session | Format |
|---|---|---|
| During village hours | HalCTF: Hostile Autonomous Layer CTF | Competition |
| During village hours | AI Village Plays Pokémon: DEF CON Edition Nick Ashworth (Maker at AI Village) | Competition and demo |
| During village hours | Cyber Mirage: Real-Time Deepfake Demos Brandon Kovacs | Live demo |
| 11:00 AM–12:00 PM | What is AI? Interactive, Unplugged Activity Sam Mosley | Interactive activity |
| 12:30–1:00 PM | Closing Remarks | Remarks |
On the DEF CON Stages · Not in the Village
| Time | Session | Stage |
|---|---|---|
| 10:00–10:30 AM | A Billion-User Blast Radius: Owning ChatGPT’s Secure Sandbox Simcha Kosman | Creator Stage 5 |
| 12:00–12:30 PM | Pwning Agentic Browsers with PleaseFix: A New Vulnerability Class for 0-Click Takeover Stav Cohen (Zenity) | Creator Stage 3 |
| 12:30–1:00 PM | Leveraging Large Language Models for Policy, Regulatory, and Compliance in IoMT: Opportunities, Risks, and Safeguards Dr. Deepti Gupta and Sai Sitharaman | Creator Stage 3 |
Poster Presentation Schedule
All 34 poster presentations are scheduled below. Poster files are added to the archive above as authors submit them.
Friday Poster Slots
| Time | Presenting |
|---|---|
| 10:00–11:00 AM | The Weight of Evidence: How an Agentic SOC Analyst Earns Your Trust Sophena Wilson Prompt Injection Testing at Scale Viviana Sutedjo and Gabriele Randi |
| 11:00 AM–12:00 PM | Poison In, Poison Out: CDC-Aware Containment for RAG and Agent Memory Kunal Jain Engineering Autonomous Security Agents for Defense at scale Dominik Swierad and Olga Shulman |
| 12:00–1:00 PM | Securing Cross-Enterprise AI Agents: An open source Identity and Governance case study Sarah Evans, Amritha Lal, Sri Aradhyula, Shankar Garikapati and Manish Singh Hacking AI: Real-World Lessons from an AI VRP John Kotheimer The Anatomy of a Chinese Knowledge Distillation Campaign Colin Shea-Blymyer and Kyle Miller |
| 1:00–2:00 PM | ARM: Agent Reasoning Markup — Detecting Sycophantic Drift and Silent Position Reversals in Multi-Agent AI Erik Roed MCParasite: Tool Poisoning and Agent-to-Agent Worm Propagation in MCP-Based AI Systems Utku Yildirim and Ozgun Kultekin Improving AI Red-Teaming by Improving Red-Teaming Reports Jessica Ji and Colin Shea-Blymyer |
| 2:00–3:00 PM | Compiling Expertise: Turning Tribal Knowledge into Auditable Agents Nathan Whitaker Memory Laundering, Pressure Points, and Other Agent Failure Modes Vincent Abruzzo, Greg Kocher, Neel Nanda and Arthur Conmy |
| 3:00–4:00 PM | The Model Is the Malware: Runtime Behavioral Detection of Malicious ML Artifacts Hala Ali and Andrew Case Poisoning the SOC: Prompt Injection via Ingested Telemetry John Seymour The Collapse of the Skill Barrier: Building Autonomous CTF Tools Through Pure Intent David Kuznicki (PuzzledHackers.org) |
| 4:00–5:00 PM | Reference Grafting in A2A: Cross-Principal Task-Data Exfiltration via referenceTaskIds Shay Sakazi, Sunders Bruskin and Emil Gelman I’ll just call you — Agent-to-Agent Privilege Boundary Failures in CI/CD Agents Dan Lisichkin |
Saturday Poster Slots
Get Involved
If you are at hacker summer camp, come by, ask questions, try the demos, and join the discussion.
- Respond to the DEF CON 34 call for volunteers to help with village participation.
- Join the AI Village community to stay connected before and after DEF CON.