Sensitive data
Health, financial, legal or employee data that you’d rather not send to a third party.
Private AI, built into the systems you already run
Some data shouldn’t leave the building. Private AI runs open-weight models on hardware in the UK or inside your own cloud account, so nothing goes to an outside AI provider. Our own research and document reading run this way.
When private makes sense
Contracts that say so
Clients or regulators who require data to stay in your environment.
High volumes
When per-call fees add up, your own hardware can cost less over a year.
Predictable costs
A fixed monthly cost instead of a bill that grows with usage.
When it doesn’t
For hard reasoning, the best hosted models are still ahead of anything you can run yourself. Often the right answer is both: private for sensitive routine work, hosted for the hard cases, with the data route written down.
You choose where the AI runs
Tier 1
A hosted model, through a UK or EU endpoint
The best quality for the least effort. Before any of your data goes to a provider, we check that its business terms rule out training on it, and we tell you which region does the processing.
Tier 2
Inside your own cloud account
The model runs in your AWS or Azure account, under your access controls and your logs.
Tier 3
A private model on hardware in the UK
Open-weight models on your hardware or ours. Nothing goes to an outside AI provider. Our own research and document reading run this way.
Storage location and processing location aren’t the same thing. We’ll tell you both, in writing.
What we set up
Models sized to the job
Open-weight models sized to the job, not the biggest available. Your test sets are re-run before any model is changed.
A key, a budget and a hard stop
A gateway where every application has its own key, its own budget and a hard stop.
Logs you control
Tracing you host yourself, with no-log routes for the most sensitive work, plus approval steps and tamper-evident logs for actions.
Screening for hidden instructions
Anything that comes from outside (web pages, emails, files) is screened for hidden instructions before a model sees it.
Live
Our own private AI platform
Our AI workflows for research, document reading and content checks run on open-weight models on our own hardware. Every call goes through a gateway with a daily budget and a kill switch.
Private is a promise you have to check
A self-hosted model is only as private as every log, backup and support path behind it. We map all of them and give you the map.
Related pages
Questions
Is it as good as ChatGPT?
For focused jobs such as sorting, extracting and summarising, it’s often close enough. We test on your examples, so you decide with numbers.
What hardware do we need?
Anything from a single workstation to a small server, depending on the job. We price it with you, or run it on UK hardware for you.
Who keeps it running?
We can, under a run plan with monitoring, updates and monthly reports. Or we hand it over to your team with documentation.
Get in touch
Tell us what can’t leave the building.
One paragraph is enough. We’ll reply with honest thoughts on whether we can help, and what it would cost.