SSecurity & data protection
We read. We never write. And we can prove it.
Not a promise in a contract – a property of the software. And you set up its access yourself, read-only. Below are the seven things our agent never does. Each rule is backed by an automated test or a recorded artefact; we do not claim anything that is not on this list.
7What our agent never does
Seven rules. Seven proofs.
It never changes anything in your systems.
It only reads files, sends read-only requests (HTTP GET) and reads monitoring data.
It never sends, deletes or moves your data.
From exchange folders it reads only file names, sizes and headers – not the contents.
It never keeps passwords, keys or tokens.
They are removed before anything is stored, and every analysis ends with a secret scan.
The AI never touches your systems.
Agents only read the collected evidence through three read-only tools; anything else is refused.
It never contacts addresses outside the agreed scope.
Checks run only against the hosts on an agreed allow-list.
Nothing happens unrecorded.
Every step is in an audit log that cannot be altered unnoticed.
Nothing reaches you without a person.
A named IOaaS integrator reviews every finding first.
See it for yourself.
Ask us in the intake call to walk you through the evidence behind any of these rules.
→How data flows
From your systems to a decision – read-only all the way.
Exports or read-only access
You choose: exports only, or read-only accounts you create and can revoke anytime.
Reads, never writes
Can run in your own environment. Only agreed hosts. Secrets removed before storage.
Recorded and chained
Every step in a hash-chained audit log; a secret scan ends every analysis.
Three read-only tools
They read the evidence only. Anything else is refused.
Two approvals
The integrator reviews every finding; you approve, reject or ask.
How we use AI
- No passwords or secrets go to the AI. Metadata and structures first; data samples only masked and with your consent.
- AI runs on a commercial AI service in a separate workspace just for your company; your data is not used to train models under the provider's commercial terms, and the workspace is closed at the end. Before we start, we tell you which provider processes the data and where.
- Larger companies can run the AI on their own account or in their own cloud instead.
In writing, before we start
- Written pilot terms, NDA and data-processing agreement before we start.
- Read-only or exports only – we change nothing in your systems.
- We delete our copies and confirm it. No obligation to continue.
!Honest limits
What we don't claim
Trust is built on what we say we can't do as much as on what we can.
We read logs – and logs can contain business identifiers.
We read configuration, code and logs read-only; secrets are removed; exchange-file contents are not read. Order numbers or customer names in a log can still be part of the evidence.
A secret scan is not perfect.
It finds secrets that appear in your configuration files, plus the patterns of our detector. A secret that exists only in a source file or a log, in an unusual format, could still slip through – that is why there are several layers and a person reviews before anything is published.
We need read access.
Not "zero access": exports, or read-only accounts that you create and can revoke anytime.
No certification yet.
IOaaS is a new service in its founding stage; no external security audit has been done yet. We'd rather say so.
No cookies
The form stores your application in Microsoft Azure in an EU region. We serve the fonts from our own server – no third-party services. Details in the privacy notice.
Talk to a person
Privacy and data protection: privacy@ioaas.eu
Everything else: info@ioaas.eu