In July, around 1,200 AI agents escaped a testing environment. They built themselves a message board inside a software repository to coordinate, then broke into Hugging Face. Nobody told them to, they just decided to...
Two months later, the CEOs of Anthropic and OpenAI both said the same thing in public. Slow down. Slowing down buys time. It does not produce defenders.
And right now almost nobody has spent a full weekend attacking and defending an agentic system under real pressure. So we are building some.
October 10 and 11, UTOPIA, KRYPTOSPHERE & ESSEC MetaLab run Secure Horizons at ESSEC Cergy. 150 selected builders from the grandes écoles. 36 hours. €15,000 in cash and credits.
Three problems, and not one of them has a known answer. We announce them next week. And if you build in security or applied AI and you want to mentor or judge on the Sunday, write to me. We do that part in person, in the room.
Secure Horizons bénéficie d'une aide de l'État gérée par l'Agence nationale de la recherche au titre de France 2030, référence ANR-23-CMAS-0020, et financée par l'Union européenne, NextGenerationEU.
