Most organizations deploy AI systems without ever testing what they'll do under adversarial input.
Intrenex is an adversarial research lab that tests how AI systems fail — from local LLM deployments to API-connected models — and publishes what we find.
What happens when you actually test
Most AI deployments pass a vendor checklist and stop there. Structured adversarial testing reveals what that checklist missed.
Adversarial Assessment V3 — Llama-3.1 8B Instruct · Mar 2026
Why most deployments ship untested
Default configurations ship vulnerable
Guardrails behave differently across model versions. The deployment guide doesn't cover what an attacker would try first.
The gap is invisible until it's not
The distance between 'we deployed a model' and 'we know how that model fails' is where real risk lives.
The attack surface exists wherever the model accepts input
Local hardware or cloud API — if it accepts natural language input, it has an attack surface that needs adversarial testing.
Controlled sandbox
Docker isolation. Ollama, PyRIT, and Promptfoo in separate containers.
Structured methodology
Threat modeling, adversarial execution, log analysis. Every test reproducible.
Framework alignment
NIST AI RMF, OWASP LLM Top 10, MITRE ATLAS.