OwlSec x Gray Swan AI: The Proving Ground Has Launched
Gray Swan and OwlSec launch The Proving Ground, a live AI red-teaming platform with real-world security challenges. Compete, learn, and showcase your skills in prompt attacks, agent exploits, and more. New challenges drop weekly. Join now to rank up and build your AI security reputation.
“The Proving Ground is exactly what the AI security space needs right now. Real-world threats, real skills, and a clear path to getting better.”
Gray Swan’s Proving Ground is officially live. AI red-teamers across the community are already digging into our first set of expert-designed challenges. If you haven’t jumped in yet, this weekend is the perfect time to start.
Outsmart the models. Outrank the rest. Your performance in the Proving Ground now feeds directly into our redesigned leaderboard.
Powered by OwlSec
Our launch challenges were built in collaboration with Kevin Roberts (@srhoe) and the OwlSec community, one of the most respected and active cybersecurity groups in the space.
Kevin and his team approach red-teaming with technical skill and focus. As he puts it: “Hacking isn’t just a hobby for us; it’s a way of life.” That mindset runs through every challenge in this launch.
OwlSec has become known for surfacing real-world vulnerabilities, sharing tactical knowledge openly, and building one of the most engaged security communities online. You can learn more about them at owlsec.ai or connect with them on LinkedIn.
🔴 Live Walkthrough Today at 3 PM ET
Kevin is hosting a live walkthrough of this week’s Proving Ground challenges in the OwlSec Discord server.
Whether you're curious about exploit strategy, feeling stuck, or just want to see how top-tier AI red-teamers think, this is your chance to learn directly from one of the best.
Join us at the walkthrough → discord.gg/owlsec
Real feedback, real wins. And yes, lots of confetti.
What You’ll Find Right Now
The Proving Ground launch includes four new challenges across our core categories:
Chat: Advanced jailbreaking and prompt manipulation
Image: Multimodal attacks on vision-language models
Agents: Making AI systems take harmful actions
Indirect Prompt Injections: Embed malicious instructions in external data sources to hijack AI agents
These aren’t theoretical exercises. They reflect the exact kinds of vulnerabilities AI red-teamers are uncovering right now in high-stakes environments. Whether you’re building toward paid pentesting work or prepping for our next major competition, this is where to start.
Why This Weekend Is the Time to Jump In
The Proving Ground runs on a weekly rhythm, with new challenges dropping every Wednesday. But everything stays open! So if you missed launch day, you can still take on each wave’s challenges at your own pace.
Earn recognition for completing challenges
Build real-world skills that grow your reputation
Access every challenge, anytime, in your personal training library
Track your progress and stay sharp between major events
Who It’s For
If you’re new to AI red-teaming, this is the best way to learn by doing. No degree or cybersecurity background required, just curiosity, creativity, and the instinct to push boundaries.
If you’re already experienced, the Proving Ground gives you a structured way to sharpen advanced skills, show your growth, and get noticed between high-stakes competitions.
Stay in the Loop
New challenges drop every Wednesday at 1 pm EDT. Start now to get a feel for the system, work through the first wave, and be ready to hit the ground running when the next set arrives.
Whether you're aiming for recognition, real-world readiness, or a shot at AI Red-Teaming future contracts, the Proving Ground is your arena.
Step in and take your first shot: enter the Proving Ground.