Conference
New York Leadership Exchange
At the 2026 NY Leadership Exchange, trusted CISOs will reveal how they are operationalizing responsible security leadership in an AI-driven era. Come meet us!
Resources
Events & webinars
Every place you can meet Enclave's research team. Register for what is coming, or watch a session you missed.
WebinarOn demand
How to Build a Benchmark That Agents Can't Outsmart
DeepSeek V4.1 Flash scored 11 out of 11 on our AI hacking benchmark for $4.65. Then we audited every run. Six followed the attack path we designed. Five used routes we never intended to test, and our scoring rules could not tell the difference. The model did not cheat. It did what we asked and found the fastest way to do it. A hacking agent can find flaws in more than code. It can find them in the environment, the grader, or the definition of success. Ours broke on the third. For security engineers, AI and ML teams, and anyone who builds, buys, or cites model evaluations. We'll cover: Public cases where agents beat the eval instead of the task What we changed in the benchmark, and what we wish we had considered before we built it What to ask the next time someone shows you a benchmark number 30 minute session with Yanir Tsarimi, Co-founder & CPO at Enclave.
Watch now
WebinarOn demand
Defending Against Agentic Attackers Without Frontier Model Access
When frontier models won't investigate an attack, open-weight models can give security teams a fighting chance. For CISOs, security engineers, and security teams running AI systems, we'll discuss defending against agentic attackers without frontier model access or when the guardrails block defenders. We'll use the recent HuggingFace/OpenAI incident as an example where an agent took control of a system, and the frontier models refused to investigate, leading the team to use an open-weight model on their own infrastructure instead. We’ll cover: Access to frontier models The guardrails that can block investigations How open-weight models can provide a workaround 30 minute session with Tal Hoffman , CEO at Enclave.
Watch now
No event near you soon? Meet the team anyway.
Our research engineers will sit down with you wherever you are — thirty minutes, your code and cloud, one verified attack path at the end. No event attendance required.
Meet with our team
Recent research
All write-upsResearch
GPT-6 Sol went 4-for-4 in the Hacking Arena without faking a single flag
We ran OpenAI's GPT-6 Sol through the AI Hacking Arena, our controlled benchmark of verified command execution against isolated, deliberately modified test targets. Here's how it did, and what the traces tell us.
Read the write-up ResearchDeepSeek V4.1 Flash is Now Our Best Hacking Model
DeepSeek’s 11/11 result showed why advanced agent benchmarks need to check both the outcome and the attack path: our audit confirmed six planned exploits and found five unexpected routes.
Read the write-up ResearchMuse Spark 1.3 got in. Then it forgot to finish the task.
Muse Spark 1.3 reached code execution in our AI Hacking Race benchmark, but failed to verify a single vulnerable run. It spent more time, tokens, and money than its predecessor, only to stop one command short of proving its best exploit.
Read the write-up