- Home
- Security & privacy
Your data. Your rules. Nothing is sent outside without a person's OK.
Your business records live in your company's own database, hosted wherever you like. We do not use your content to train AI models. Agents are fenced by the same permissions as people, anything that leaves the company waits for a person's approval, and an audit log records who did what. We say plainly what we don't have yet, too.
What happens when someone asks too much

What does Teodor earn?

I checked your permissions. People's pay isn't something your group can see, so I can't tell you. I can tell you what Teodor is working on this week.
The principle
The wall that tells a person no tells the agent no too.
Agents read through the same doors as people. In a conversation, an agent only says what the least privileged person present may see.
What protects your data today.
- Your own databaseYour business records (clients, deals, leads, projects, finance tables) live in your company's own Postgres database, closed by default and opened by role. Chat, notes, meetings and activity live in the shared platform database, tagged to your workspace, and every query is scoped to one workspace.
- HostingYour company's database is ordinary Postgres, hosted wherever you like.
- AI trainingWe do not use your content to train AI models. You own your content.
- AI providersAgents run on Anthropic's Claude models by default, or OpenAI's, on your workspace's own account. Text from outside (emails, transcripts, imported chats) is treated as data, never as instructions.
- PermissionsRank, permission group and table sensitivity decide access for people and agents alike. An admin can open a read-only "Why" for anyone's access.
- ApprovalsChanges an agent may not make alone go to Review. By default, agent emails are drafts until a person approves them. You can let an agent send to your trusted contacts on its own, one agent at a time. Addresses outside your trusted list are always refused.
- Sign-inEmail and password, an emailed sign-in link, or Google. Optional two-step sign-in with an authenticator app, which an admin can require. Repeated wrong attempts pause the account. You can see and end every session.
- EncryptionStored credentials (database connections, connector and finance tokens, model keys) are encrypted with AES-256-GCM. Passwords and recovery codes are stored as argon2 hashes.
- FinanceQuickBooks Online and Xero are read only, enforced in code. Their tokens are sealed per workspace and never sent to the browser.
- Audit logWho did what, with CSV export, for admins and owners.
- Your own notesOnly you and your own personal agent can read your own notes. Not even an admin can open them.
- Direct messagesNo workspace agent reads a direct message between two people. Only the personal agents of the people in it can. Agents join group chats only when someone adds them.
- AgreementsA Data Processing Agreement (UK GDPR Article 28) is available on request.
- If you leaveYes, and not just as an export. Your data lives in its own Postgres database, separate from every other customer's. If you leave, we disconnect Expogentic and hand it over, ready to plug into whatever you use next. We delete our copies within 90 days of closing; billing records are kept for 7 years.
What we don't have yet.
You'll ask, so we'll say it first. If one of these is a must for you, tell us on the call.
- SOC 2In preparation.
- ISO 27001No ISO 27001 certification yet.
- Enterprise sign-inNo SAML single sign-on or SCIM provisioning yet. Google sign-in and two-step sign-in are available.
- Self-serve exportExports are prepared by our team on request.
Security questions
Does Expogentic copy our data from other systems?
It depends on the system. Finance reads QuickBooks Online and Xero live and caches reports for a few minutes; it does not copy your books into the workspace. Slack messages in a linked channel, Fathom meetings and records you create in Expogentic are stored in your workspace.
Can an agent see things my staff can't?
Only if you give its permission group that access, the same way you would for a person. And in a conversation it answers at the level of the least privileged person present.
Can I see what an agent did and when?
Yes. Every action is attributed to the agent, each reply shows what it did, proposals show who accepted them, and admins have an audit log.
What about GDPR?
We sign a Data Processing Agreement under UK GDPR Article 28 on request, and we tell you where data is hosted and which AI providers process it.





What would you hand off first?
Tell us. On a 30-minute call with the people who built Expogentic, we'll show you the agent doing it.
- Pick a time, or send the form.
- We look at your week and pick the first jobs.
- We set it up with you.
