The system runs in your environment. Your data stays under your control.
Security for an applied AI system comes down to four questions: where it runs, what happens to your data, who owns what, and what is recorded. This page answers all four the way an Applicat AI engagement answers them, in the commitments that go into the contract. The company was founded in 2026 and holds no security certification, so there is no badge at the top of this page and the answers have to do the work instead.
Where it runs.
Into your own environment, on your own accounts, behind the identity provider you already run. The practical consequence is the one a reviewer cares about: the controls your organisation has already built are the controls the system sits inside.
Your cloud, your accounts
Systems are deployed into your own Microsoft Azure, AWS or Google Cloud environment, or into a dedicated one under your control. The workload sits in your tenancy, and if you end the relationship nothing has to be moved.
Behind your identity provider
Access follows the identity provider you already run. Who may use a system, and who may see what inside it, stays a decision your own directory makes and can reverse.
The proof runs there too
The first four to six weeks are built on your real data, in your own environment, with the permissions that genuinely apply. There is no phase on sample data outside your perimeter, so there is no later migration into it.
Where data cannot cross a border, it does not
Residency, retention and sovereignty requirements are settled at the Frame stage, before any architecture is fixed, and they drive the choice of model, region and architecture from there on. Where nothing may leave a jurisdiction, the system is built on open-weight models running inside it.
What happens to your data.
Nowhere it is not already. Model access is the one point at which anything leaves your environment, and the terms on that path are the first commitment below.
01Nothing of yours trains a model
Model access goes through enterprise endpoints where your data is not used for training, and nothing of yours goes into training a shared model.
02Your process knowledge does not travel
We commit in writing that what we learn about how your business works stays with your business, and that we will not reuse your process knowledge on another client.
03Retrieval inherits your permissions
Search across your own material inherits the permissions of the system the material came from, so nobody sees what their login would not show them, and every answer carries a citation back to the source record.
04Least access, and credentials held once
Every system gets the least access it needs and a fixed set of tools it is allowed to use, scoped to a task and bounded in time. Credentials are brokered once and never embedded in an application, so what a system can reach stays one decision your security team makes, and can revoke.
That covers the systems we build. Personal data collected by this website and by our own business relationships is a separate matter, and the privacy policy is the statement of it: what we collect, the lawful basis, the processors, transfers outside the United Kingdom under the UK International Data Transfer Agreement or adequacy regulations, enquiry data kept for up to 24 months, and your rights under UK GDPR.
Read the privacy policyWho owns what.
Ownership is a test a reviewer can apply: could your team, or another firm, pick this up and carry it on? From the first sprint, the answer is meant to be yes.
- The code and the infrastructure
- Repositories, infrastructure as code and configuration, in your accounts and under your source control from the first week.
- The data, the prompts, the tests and the baselines
- Yours in the contract. The cases, the scoring and the baselines are what let you judge the next model, or the next supplier, without taking anyone at their word.
- No licence of ours in the middle
- Systems are assembled from standard and open components. There is no Applicat AI runtime to license, no fee per seat, and nothing that stops working when a contract with us ends.
- The handover pack is a deliverable
- It is produced during the Build stage, while the people who built the system are still on it. End the relationship tomorrow and the system keeps running, with your team holding everything they need to change it.
One honest limit. The models, and most of the systems a build connects to, are other people's software, so independence here means you can swap any of them.
What is recorded, and who can read it.
One trail in one format, written as the system is built, while it is still cheap to write.
One trail across everything that runs
What was asked, what was retrieved, which tools were called, which model answered and who approved it. Run-level logging covers inputs, retrieval, tool calls, model and cost.
The decisions a person took
A person stays in the loop wherever an action is hard to undo, and those decisions are recorded at the checkpoints where they are required. Every action an agent takes is logged, attributable and reversible.
One export that answers all three
Internal audit, a customer due-diligence questionnaire and the incoming AI regulations want the same evidence out of the system. Every system also keeps its test results and its change log.
What Applicat AI cannot show you.
A security page is the worst place on a website to be vague, so here is the part a competitor fills with badges and logos.
01
No certification, and no claim of one
Applicat AI was founded in 2026. It holds no ISO 27001 certificate, no SOC 2 report and no equivalent attestation, and this page will not describe an audit that has not happened. When that changes it will be stated here before it is stated anywhere else.
02
No named clients yet
We publish no case studies, because there are none a client has yet agreed to be named in. In place of a logo wall we will introduce you to a reference directly, which tells you more than a paragraph we wrote about ourselves.
03
What you can check today instead
That the company is founder-owned, with no AI lab, cloud provider, consultancy or investment fund on the cap table: ask for the register and we will send it. That the engineers named on the engagement are the ones who do the work. And that a first engagement can end with a written recommendation against building at all.
What your security review receives, and when.
All of it is written up during the Build stage, before release. Your reviewers get something complete to approve or reject, and nobody is asked to sign off an intention.
- Identity and access: who and what may reach the system, with which permissions, and where those permissions are changed
- Data handling: where data sits, where it moves, and what the model endpoints do with it
- Logging: what is recorded on every run, and how to get it out
- What the system may and may not do, and the points where it has to stop and ask a person
- The due diligence on every vendor in the chain
- The test set that every future change has to pass before it reaches a user
FAQ
Questions a security review asks
Send this page to whoever reviews it.
If an answer here is not the answer your organisation needs, tell us which one and we will say plainly whether we can meet it.