Security
Your data stays yours.
You’re letting software into the middle of your business, so you should know exactly how it’s looked after. Here’s what goes into a build, in plain English. Your IT people can ask us anything.
- 0Security incidents on client builds
- 100%Data encrypted
- 148Point security checklist
- 100%Agent actions logged
Four rules we build by
Built in from day one
Who can see what gets decided before the first login screen is written. Security bolted on later is how things get missed.
More than one lock
Encryption, access by role and each client’s data kept apart. If one control fails, the next one’s still standing.
Tested, not assumed
Each control gets tested before launch, and a separate review tries to break the code before it ships.
Straight answers
We’ll fill in your security questionnaire, walk your IT people through the build and tell you about any gap we find.
What we check
Every build goes through our 148-point production and security checklist before it goes live. Not every build needs every line. We’ll show you which apply to yours, and the proof.
Your data
- Encryption
- TLS in transit, AES-256 at rest on our standard hosting
- Where it runs
- Accounts in your name, in Australia where the services allow
- AI training
- None. Claude’s commercial API doesn’t train on your inputs
- Separation
- Each client’s data kept apart, and tested
- Backups
- Confirmed, with a restore drill before launch
Who gets in
- Staff sign-in
- Two-factor
- Permissions
- By role, checked on the server
- Passwords
- 12 characters or more, checked against known leaks
- Login attacks
- Rate limits and bot checks
- Audit trail
- Logins, role changes and data exports all logged
The build
- Code review
- Independent review before anything merges
- Secrets
- Key store only, scanned out of code and history
- Inputs
- Validated at every boundary
- File uploads
- Type and size checked, private storage
- Dependencies
- Audited before launch
The AI
- Approval
- A person approves anything that leaves the business
- Access
- Agents get the least access they need
- Your data only
- It only ever sees your own data
- Prompt injection
- Tested on anything an outsider can write
- AI output
- Treated as untrusted, never run as code
Keeping it running
- Alerts
- Every live system names who gets the alert
- Recovery
- A written plan for how it comes back, and how long it takes
Between us
- NDA
- Signed as standard
- Ownership
- You own it. It runs in accounts in your name
- Handover
- Documentation and training included
- If we part ways
- It keeps running
Next step
The infrastructure gap is costing you more than you think.
The businesses pulling ahead aren’t working harder. They’re running on better infrastructure. We build it, you own it, and it compounds. 15 minutes with a founder, no slides. If there’s no saving worth chasing, we’ll tell you.
Questions, answered first
What does Aether AI actually build?
Whatever your audit says will pay. We build the infrastructure that lets your business run without you in every loop: AI agents, workflow automation, data pipelines, live dashboards and custom software. If you can explain the work, we can automate it, even if it’s still on paper today. It might be one agent, or a platform that runs the whole business.
How long does it take?
The audit takes three to four weeks. Most first builds go live in weeks, not months.
Do we have to change our software?
You don’t have to. We plug into what you already run, like Xero, OneDrive, your inboxes and email. If the audit shows that having everything in one place would be cheaper and easier, we’ll show you the numbers and you decide. Our job is the easiest implementation: one that doesn’t disrupt how your team works, only makes it faster, gives your staff their time back and gives you room to scale.
Is it safe to let AI into my business?
Yes, and here’s how we keep it that way. It runs in accounts in your name and starts read-only. A person approves anything that leaves the business, and every step is logged. It’s built on Claude’s commercial API, which doesn’t train on your inputs. Every build goes through our 148-point production and security checklist before it goes live.