DEF CON asked how it works. Black Hat asked what it proves. Everyone asked: who sees the data? That question gets to the heart of AI pentesting: trust. Our answer: ✅ frontier LLMs, run in the US ✅ we

DEF CON asked how it works. Black Hat asked what it proves. Everyone asked: who sees the data? That question gets to the heart of AI pentesting: trust. Our answer: ✅ frontier LLMs, run in the US ✅ we validate exploits, not flag guesses ✅ a decade of offensive security work, not a prompt wrapped in a UI Robert Tanase and Jan Pedersen unpack what we learned at DEF CON and Black Hat, and what's next