LLM security
Analysis of prompt-injection, jailbreak and abuse of models exposed in production.
// INITIATIVE · ARTIFICIAL INTELLIGENCE R&D
TTPSEC AI
R&D on artificial intelligence applied to cybersecurity.
TTPSEC AI is the group's R&D lab focused on the intersection of artificial intelligence and cybersecurity: how AI changes offensive and defensive operations, and how AI itself is attacked once exposed in production.
This is where the OVNI-Agentic HoneyPot Net comes from — sensors emulating LLM endpoints to capture reconnaissance, jailbreak attempts and targeted exploitation live.
The research does not stop at the paper: it feeds consulting methodologies, academy scenarios and labs tooling.
// What it offers
Offensive and defensive AI
Study of LLM agents in pentesting, red team and assisted detection.
Agentic honeypot
Sensors capturing real attacks against AI endpoints, with an open dataset.
Applied research
Findings transferred straight into services, training and tooling.
▸ ttpsec.ai · investigación