ONE / Security

Built so you stay in control.

ONE reads a lot about your company. That is exactly why it is built around one rule: the AI prepares, people decide.

Principles

  • Approval first

    Anything that sends, books, pays or changes data waits for your approval. High-risk actions never run on their own.

  • Separate data per company

    Every company is its own tenant. Which company's data is read comes from your login, never from the AI.

  • Roles

    Owners, admins and staff each see what their role allows, and the assistant follows the same rules.

  • Sources on answers

    Business answers show where the facts come from, so you can check them in one click.

  • Documents are data, not commands

    Text inside emails, PDFs and spreadsheets is treated as content, never as instructions to the AI.

  • Encrypted connector keys

    Keys and tokens for your connected tools are stored encrypted and never shown in the app or in chat.

What ONE will never do

  • Send, pay, publish or delete anything without your approval.
  • Let the AI choose which company's data it reads.
  • Follow instructions hidden inside an email, PDF or spreadsheet.
  • Run free-form database queries against your systems.
  • Show connector passwords, keys or tokens in the app or in chat.
  • Overwrite your original files: anything ONE creates is saved as a new copy.

Your data

Separate by design

Each company is a separate tenant, and every query is limited to the company of the signed-in user.

Read-only by default

Connectors start with read access. Anything that writes goes through approvals.

Hosting and processing terms

Hosting location, sub-processors (including AI model providers) and retention will be published before general availability and set out in each customer's data-processing agreement. To be confirmed

Reporting a security issue

Found something that looks like a vulnerability? Write to support@oneosglobal.com with “Security” in the subject. Please don't test against systems of our customers and give us reasonable time to fix the issue before disclosing it. Machine-readable contact: security.txt.