Hire me NL
AI Offensive AI

Your AI is a new way in. We attack it before someone else does.

Chatbots, agents and AI tools now read your mail, your documents and your systems. BlockTheChain attacks them the way a real attacker would, puts AI to work in your own authorised testing, and builds private AI that keeps your data yours.

Only systems you own or are authorised to test, with written permission first.

01 What we do

Three services. Attack, accelerate, build.

Test the AI you already run, speed up your own testing with it, or get a private setup that stays under your control.

Attack

AI red teaming

We attack your chatbots, agents and connected tools: hidden instructions, poisoned tools, data leaks and bypassed safety rules. You get the evidence and the fixes.

Accelerate

Offensive AI for your team

AI that speeds up your own authorised testing: mapping exposure, finding attack paths and chaining findings. It runs on private infrastructure.

Build

Secure private AI

One controlled AI gateway for your staff, with logging and access control. No company data is used to train outside models.

02 What you get

Evidence, not a demo. Then it keeps working without us.

Three deliverables, each one usable by your own team after hand-over. No slide deck, no dependency on us.

Report

Attack report with evidence

Every finding with the input, the response and the impact, ranked by real risk. Then a retest.

Playbooks

AI playbooks your team can rerun

The AI-assisted steps, documented and handed over. No black box.

Platform

A private AI setup

Deployed on infrastructure you control, with the access rules and logs your auditor will ask for.

03 Proof

We run our own AI platform. Built here, live in production.

Warface Protocol is the group BlockTheChain belongs to, and its AI platform. Gateway, routing, access control and logging: the same approach we bring to your company.

Two of the free labs come from this work: the hidden-instruction scanner and the MCP tool-poisoning scanner.

04 The line we keep

Offensive work. Clear rules.

Four rules that apply to every AI engagement, written down before the first test. They do not change per client.

  1. Permission first. We only test systems you own or are authorised to test, and only with written permission.
  2. Your data stays yours. Nothing you share is used to train any model.
  3. No black box. Every AI-assisted step is documented, so your team can check it and rerun it.
  4. No lock-in. We tell you which parts you can run without us.
05 Start

One message starts it.

Thirty minutes to scope it. A fixed-price proposal within two business days.

Goes straight to the operator. Nothing is stored on this site.