How we keep your data safe.
Your AI runs on your servers or in your region. Your data is never used to train models.
In short
Attentions is ISO 27001 certified, and our SOC 2 Type I report is available on request. Your AI runs on your servers or in your region. Data is encrypted in transit and at rest. It is never used to train models.
- Your servers or your region
- Keys you can hold
- No training on your data
- Every step logged
Our certifications and compliance.
Your security team can review each one. Ask us for the reports.
ISO 27001 certified
Our information security management is certified to ISO 27001, the international standard.
SOC 2 Type I report (available on request)
An independent auditor's report on how our security controls are designed. Ask us for a copy.
Request the reportGDPR compliant
We handle personal data under the EU's data protection rules, including your rights to access and deletion.
HIPAA compliant
Health data in the US is handled under the HIPAA privacy and security rules.
Controls built into every layer.
Who can see what, how your data is protected, and what the AI may do.
01
Who can see what
Every answer respects who is asking.
Company sign-on
Sign in with your company login (SSO). Accounts follow your company directory (SCIM).
Roles and attributes
Access by role and by attributes, such as team (RBAC and ABAC). Every request carries the person who asked.
Permissions first
Permissions are checked before anything is searched. If you cannot open a file today, the AI cannot show it.
Access reviews
Who has access is reviewed regularly, as part of how we run the platform.
02
How data is protected
Encrypted, with keys you can hold, and never used for training.
Encryption
Data is encrypted in transit and at rest.
Customer-managed keys
You can hold the encryption keys yourself, so access to your data stays in your control.
No training on your data
Your data is never used to train models.
Leak checks
Leak checks (DLP) hide personal data in answers. Retention rules set how long data is kept.
03
What the AI may do
Allowed steps only, with a person where it matters.
Allowed steps only
Each AI worker can use only the actions you approve. Checks on the way in block hidden instructions.
People approve
A named person approves risky steps before they happen. Nothing changes a system without an approver.
Stop switch
A person can stop any run at once.
Audit logs
Every step is logged: the data, tools, model version and approver. Logs cannot be changed, and runs can be replayed.
Your servers or your region. You choose.
You decide where the AI Platform runs. Your data stays where your rules say it must.
Three ways to run it
On your servers.
In your data centre or your own cloud account. Nothing leaves your network.
In your region.
Data stays in your region, for your data residency rules. You hold the keys.
With an approved cloud model.
Used only if you approve it, and only for the data you allow.
How we handle your data
Read the privacy notice- During the AI assessment
- Read-only access, approved by your security owner before Day 1. Analysis runs inside your environment. Results are shown by system and team, never by person.
- Where data stays
- On your servers, or in your country or region.
- Training
- Never used to train models.
- Retention
- Retention rules set how long data is kept. Memory can be deleted.
- This website
- The site does not store your intake answers. They go to us by email or to Calendly.
Every change is checked before it ships.
We plan, build, package, deploy and run the AI Platform with security checks at each step.
01
Plan
A threat model, data classification and an architecture review, before we build.02
Build
Secure coding. Automatic scans check our code (SAST), the libraries it uses and any leaked secrets.03
Package
Hardened images and signed builds, with a record of where each was built. An SBOM lists every software part inside.04
Deploy
Set up as code and approved by named owners. Releases go out in small, watched steps (canary or blue-green).05
Operate
Security log monitoring (SIEM), fixes for known weaknesses, access reviews and incident response.
Rollback
Any change can be rolled back to the last good version: the prompt, the model or the search index.
Approved models only
Models come from an approved list. Versions are tracked and files are signed.
Incident response
Every failure has a planned answer. If a release goes wrong: a guardrail stop, a rollback and an incident review.
The papers your reviewers ask for.
We share these with your security and risk teams.
SOC 2 Type I report
An independent auditor's report on how our controls are designed.
ISO 27001 certificate
Proof that our information security management is certified.
AI risk papers
An AI inventory, a risk class and a control matrix, for your risk review.
Report a security issue
Found a weakness? Tell us.
Email our security contact with the details.
- Security contact
- puneet@attentions.ai
Please include
- What you found, and where.
- The steps to see it again.
- How we can reach you.
Please give us time to fix it before you share it publicly.
Email the security contactQuestions security teams ask us.
Short answers. Ask our AI experts for the detail.
Is Attentions ISO 27001 certified?
Yes. Attentions is ISO 27001 certified. Ask us for a copy of the certificate.
Do you have a SOC 2 Type I report?
Yes. Our SOC 2 Type I report is available on request. Email us and we will share it with your security team.
Are you GDPR and HIPAA compliant?
Yes. We are GDPR compliant and HIPAA compliant. Ask us for the details your reviewers need.
Does our data leave our network?
Not unless you choose. The AI Platform runs on your servers or in your region. An approved cloud model sees only the data you allow.
Is our data used to train models?
No. Your data is never used to train models.
Who holds the encryption keys?
You can. Data is encrypted in transit and at rest. With customer-managed keys, the keys stay with you.
How do I report a security issue?
Email puneet@attentions.ai with what you found, where, and the steps to see it. Please give us time to fix it before you share it.
Bring your security team to the first call.
Our AI experts will walk them through every control on this page.
It starts with a short call with our AI experts.