AI statement

Where artificial intelligence was used to make JDX Stronghold, where it runs inside it, and what it cannot be relied on to do.

How the software was built

JDX Stronghold is designed, directed and owned by JDX Security. A large part of its source code was written with the help of an AI coding assistant (Claude, made by Anthropic), working to JDX Security's instructions.

Every change is reviewed, built and tested before release, against an automated test suite and a security checklist covering sign-in, encryption, permissions and network exposure. No customer footage, customer data or site information is given to an AI service to build or test the software.

JDX Security is responsible for the software it sells, however each line of it was written. The AI tool is not a party to your licence and makes no promises to you.

AI inside the product

Object detection (optional; it can be declined during installation). An open-source model (YOLOX, Apache 2.0) runs on your own server and labels people, vehicles and animals in the picture, so a zone can alarm on a person rather than on any movement.
Footage search (optional). A model on your own server describes recorded frames so similar moments can be found again.
Licence plates. Plates are read by the plate camera's own recognition. JDX Stronghold stores the reads and can act on them by your rules, for example opening a gate for a plate on your allow list.

All of this runs on your server. Footage, images and plate reads are not sent to JDX Security or to any AI provider. JDX Stronghold does not do face recognition.

What it cannot be relied on to do

Detection and plate reading can be wrong: they can miss something that is there, or report something that is not. Lighting, weather, angle, distance, dirt and a camera's settings all affect them.
They support a person watching the site; they do not replace one. Do not rely on them alone where a miss or a false alarm could cause harm.
Automated actions (such as opening a gate for a plate) have safety checks - an exact match, a confidence threshold, a cooldown and a log of every action - but you choose the rules and are responsible for them.

Your responsibilities

Using cameras, plate reading and automated decisions about people is governed by privacy and surveillance law where your site is (in New Zealand, the Privacy Act 2020 and the Privacy Commissioner's guidance on CCTV and number plate recognition). You are responsible for signage, notices, retention and any consent needed. See clauses 9 and 18 of the licence agreement.

Questions

Ask at [email protected].