Sociail · Trust
Commitments and controls
Control, proof, approval

Privacy-first AI.

Sociail is built for shared AI work with memory you can inspect, actions you approve, and automation that stays inside visible boundaries.

WithoutInvisible memoryWithoutHidden actionsWithoutUncontrolled automation
Read the Backbone
AI participation trust traits

Three traits carry the trust model.

Privacy-first AI is an operating model: bounded action, scoped room context, and explicit role boundaries that users can inspect before they rely on AI.

Approval paths

Hands

Hands stay bounded by grants, approval cards, and receipts before meaningful action crosses a boundary.

Scoped memory

Memory

Memory stays room-scoped by default, visible to users, correctable, revocable, and tied back to source context.

Role boundaries

Soul

Soul makes the AI teammate identity and boundaries explicit so users can understand why it behaves the way it does.

01The trust ledger

What Sociail stores —
and what it doesn't.

What we store

Sociail does store

  • +
    Your room content
    Messages, artifacts, decisions inside rooms you own.
  • +
    Your AI prompts and responses
    So you can scroll back, audit, and reuse.
  • +
    Receipts on AI actions
    What happened, when, under whose authority.
  • +
    Account & workspace info
    Name, email, billing — what's needed to run the service.
  • +
    Browser Control approvals
    Which page, which action, when you said yes.
What we don't

Sociail does not do

  • ×
    Train external models on your content
    Not by default. Workspace content is not used to train general AI or foundation models unless you explicitly opt in or sign a written agreement.
  • ×
    Sell your data
    To anyone. Ever. There's no business model behind us doing this.
  • ×
    Mix data across workspaces
    One team's context never informs another's AI.
  • ×
    Run AI in the background unprompted
    If it's consequential, it's an approval card — not a silent action.
  • ×
    Read browser pages you didn't approve
    The extension acts on the page you said yes to. Nothing else.
02How memory works

Room context is scoped by default.

AI can use the room context you make available: messages, artifacts, and decisions inside that room. It does not automatically know every room in a workspace, and it never knows other workspaces.

Broader continuity, cross-room access, and long-term memory are explicit, visible, and governed rather than hidden or automatic.

There's no hidden “AI memory” silently collecting context outside the rooms it's in. If we're remembering something, you can find where it came from.

03How approvals work

Every consequential action waits for a human.

AI proposes; you approve, edit, or reject. The approval card is visible in the room — the people doing the work see it, the decision is recorded, Sociail is designed to route consequential actions through the approval path.

Approval requirements are configurable per AI teammate, per room, and per category of action. Low-stakes actions can run without approval if you choose. Consequential actions are designed to stay on the approval path unless a workspace explicitly grants otherwise.

We don't think of approvals as friction. We think of them as the contract that makes AI-assisted work trustworthy.

See Trust & Approvals
04How Browser Control is bound

A concrete trust commitment users can inspect.

Browser Control is bounded by three rules — together, they're what makes a browser-acting AI teammate something you can actually trust:

1. User-enabled. The extension is installed and turned on by you. We don't silently activate.
2. Room-scoped. Browser Control activates from a specific Sociail room — not from the browser tab.
3. Per-action approved. Browser actions require per-action approval. Sociail does not provide a background “approve once, run forever” mode.

Sociail does not read browser pages in the background. Sociail does not use the extension to monitor your browsing. The extension only acts when you've approved a specific action — and you can see exactly what it did, after.

See Browser Control
05How AI providers are used

We don't pretend the AI runs in our basement.

Sociail uses third-party AI model providers to power AI features. We are transparent about this because provider use is part of the real trust model, not a footnote.

When you direct a prompt to an AI teammate:

— The necessary inputs are sent to the provider for that request, and only that request.
— The provider processes the request and returns a response.
— External provider routes must pass policy, retention, and training review.
— We don't send more than the request needs to fulfill itself.

We don't pretend providers see nothing. We minimize what is sent, and provider handling remains governed by applicable service terms, retention limits, and data-processing controls.

06How you delete, revoke, or correct

You stay in control of what AI knows.

Trust without an exit is a trap. The controls below are designed so you can change your mind — at any layer — without having to ask permission.

Delete a room
Its content is removed from AI memory. AI teammates that operated in that room lose the context. Receipts of past actions remain in audit, scrubbed of room content.
Delete an artifact
It's removed. Subsequent AI runs no longer have it as context.
Revoke an AI teammate
The AI teammate is gone. Action history stays for audit. It cannot run again unless re-invited.
Correct a receipt
Edit or annotate. Corrections are tracked alongside the original — no rewriting history.
Export your data
Request via workspace settings (or through Help during Early Access). You get your content in a portable format.
Delete your workspace
All content removed within a reasonable period — except where retention is required by law.
07Early Access limitations

Honest about what's still being built.

Heads up

A few trust controls aren't fully self-serve yet.

We'd rather tell you now than have your team discover the gap. If any of these are critical for your team today, tell us before onboarding.

  • Self-serve data export is in development. For now, request it through Help and we'll deliver it manually.
  • Provider list in workspace settings is finalized but not yet displayed in-product. Ask if you need the current list before signing up.
  • Per-action audit logs are being expanded. Current logs are room-scoped and visible in the room itself.
  • Per-room retention controls are on the roadmap. Workspace-level retention is in place; granular controls are coming.

We will keep this page current as controls move from assisted to self-serve.

See trust in a workflow
Review-first research with approval gates
A concrete workflow example where source checks, Browser Control, and human review stay visible.
/how-it-works/competitive-sweep
/backboneThe technical foundation

Trust grammar that's enforced, not promised.

Everything on this page is operational because the architecture beneath it makes it operational. Eight layers — SOUL, Cortex, Context Firewall, Policy & Grants, Memory & Data Fabric, Tool Broker, Receipts, Work Graph — that turn each commitment above into something the system can verify, not just a promise on this page.

Read the architecture
The Backbone
Eight layers behind the trust commitments. The page where the words above become enforceable.
/backbone
Legal companions

Plain-language trust, backed by the formal documents.

This page explains how Sociail thinks about trust in practice. The Privacy Policy covers data handling, and the Terms of Use cover the rules for using the service.

Read Privacy PolicyRead Terms of Use
Trust-first Early Access
AI you can actually bring into the work.

Sociail is in request-led Early Access. We're working with small teams ready for AI inside shared work — with the trust grammar in place from day one.

How Early Access works