Security & data
For banks, pharmaceutical and manufacturing companies, the question is not what AI can do. It is where the data lives, who can access it and who is accountable for decisions. Here is how we build solutions to meet your security requirements, and what exactly we agree in writing at the start of a project.
Where your data lives: deployment options
An AI solution can be deployed inside your own infrastructure: in your environment, on your hardware, on-premise. In that setup, data never leaves for an external cloud, and processing happens where you are in control.
We choose the deployment model for your case: fully inside your environment, in your private cloud, or a hybrid. The boundary data never crosses is defined and documented at the start of the project, so your security team can see it before the first line of code is written.
Which models we use
We pick the model to fit the task, not the task to fit the model. For sensitive data, we use open-source models that run locally on your hardware, with no requests sent to external services. Where it makes sense, we can also use external models, but that is always a separate decision agreed with you.
We do not lock you into a particular vendor or default to one. Choosing the model is part of the project, not a fixed stack carried over from the previous client.
Access and encryption
Access to the system and the data is role-based: everyone sees only what they need for their work. Actions on data are logged, so there is a clear trail of who did what and when. Data is protected in transit and at rest.
We agree the specific security controls with your security team, to fit your infrastructure and policies. We do not impose our own standard on top of yours; we fit into what already works for you.
GDPR and legal terms
We design solutions to meet GDPR requirements: data minimization, a defined legal basis for processing, and the ability to delete data. On request, we sign an NDA and data processing terms that set out what is processed and how.
This is our commitment to work by your rules, not a substitute for your own legal assessment. The final call on compliance always rests with your lawyers and security team.
Human in the loop, and what happens if the agent gets it wrong
We never hand critical decisions entirely to an agent. Where a mistake is costly, a human stays in the loop: the agent prepares an option, a person approves it. For every scenario, we decide in advance what the agent does on its own and what it escalates for approval.
If the agent is unsure or steps outside what it is allowed to do, it stops and hands the case to a person instead of guessing. The agent’s permissions and the escalation path are written down before launch, so you know in advance how the system behaves in a disputed case instead of finding out in production.
Computer vision and privacy
Our computer vision is adapted to EU law. We do not collect faces or build personal data about people. The system works with commercial metrics: shelf fill, merchandising, and sales on the shop floor.
In other words, you get analytics on products and space, not surveillance of visitors. For retail, this removes a whole class of legal risk at the level of what the camera records in the first place.
Support and SLA
Support and response terms are set out in the contract: what support covers, response times, and how updates are rolled out. Support starts from $150 per month, and the scope depends on how complex the solution is. We have already delivered large projects with heightened security requirements, so we are ready to work by your procedures rather than impose ours.
FAQ
Frequently asked questions
Can we deploy the solution entirely in our own infrastructure?
Yes. The solution can be deployed on-premise, in your environment, on your hardware, with no data leaving for an external cloud.
Is our data sent to external cloud services?
Not with local models. The boundary data never crosses is agreed and documented at the start of the project.
Which models are used for sensitive data?
Open-source models that run locally on your hardware. We use external models only when separately agreed with you.
Do you sign NDAs?
Yes. On request, we sign an NDA and data processing terms that set out what is processed and how.
Who is responsible if the agent makes a mistake?
A person approves critical decisions. We write down the agent’s permissions and the escalation path before launch, so the system does not guess in disputed cases.
Does your computer vision collect people’s faces?
No. We work with merchandising, shelves and sales on the shop floor, without faces or personal data about people.
Free consultation
Let us show you how it would look in your environment
Request a free consultation. We will go through your infrastructure, security requirements and deployment options, so you can see the data boundary and how the solution will work before the project starts.
What you'll get from the consultation