Security & Governance
A boundary enforced where the data lives.
SecureShield AI is a sales and revenue intelligence platform. Its governance is part of how it is built: each workspace is isolated in the database, access depends on role, and files, calls and conversations belong to the workspace that created them. This page describes only what is implemented.
Isolation
Every query is scoped to your organization through Row Level Security in the database. The checks run in PostgreSQL on every read and write, not only in the screens.
Access
Owners, admins and members work in the same system with the appropriate level of control. Role checks are made by the database.
Ownership
Files, calls and conversations belong to your workspace and are not mixed with another organization's data. SecureShield AI does not train models on them.
How isolation works
Row Level Security, on every customer table.
Row Level Security (RLS) is a PostgreSQL feature that attaches a rule to a table. Whenever anyone queries that table, the database applies the rule to every row. SecureShield AI's rule for customer data is simple: a row is visible only if its organization is one the signed-in user belongs to.
Tables covered
- accounts
- contacts
- leads
- opportunities
- opportunity stage history
- sales activities
- conversations
- conversation analyses
- briefings
- actions
- insights
- reports
- assistant history
- data imports
- activity log
- credit balances and usage
- subscriptions
Organization membership, invitations and profiles are also protected by policy; a user can only see people who share a workspace with them.
What this means in practice
- A user in Organization A cannot read, change or delete Organization B's rows — even by calling the database API directly.
- The AI Assistant and every AI feature load data through the signed-in user's session, so they only ever see that user's workspace.
- Hiding a button is never the only control. If the interface were bypassed, the database would still refuse.
The repository includes an automated test that applies the migrations to a database and verifies these properties, including cross-organization reads, writes, deletes and file access.
Access control
Who sees what.
Three roles. The table below is generated from the permission list used by the application; the right-hand column names the mechanism that enforces each row.
| Capability | Owner | Admin | Member | Enforced by |
|---|---|---|---|---|
| View all records in the workspace | RLS select policies | |||
| Create and edit accounts, leads, opportunities and activities | RLS insert/update policies | |||
| Add conversations and call recordings, run AI analysis | RLS + storage policies; credit RPC | |||
| Generate briefings, ask the AI Assistant, generate reports | RLS + credit RPC | |||
| Work the action queue (complete, dismiss, add) | RLS update policy | |||
| Delete records, conversations and recordings | RLS delete policies | |||
| Invite members and remove members | create_invitation / remove_member RPCs | |||
| Edit workspace details | RLS update policy on organizations | |||
| Invite or remove admins and change member roles | set_member_role / create_invitation RPCs |
Members join by invitation tied to a specific email address. The owner role cannot be removed or reassigned in the application.
Your data and AI
What happens when you run an AI feature.
Briefings, conversation analysis, drafted follow-ups and AI Assistant answers are produced by Google's Gemini API. The request is made from SecureShield AI's server, never from your browser, and includes only the workspace data needed for that request.
SecureShield AI does not train or fine-tune models on customer data and does not pool data across workspaces. Briefings store the exact figures they were written from, so each statement can be checked against raw data.
AI output is labelled as AI-generated everywhere it appears, and is always separate from the records themselves.
Infrastructure used
- Supabase
- Authentication, PostgreSQL database and private file storage.
- Google Gemini API
- Generates AI output for the features above, when you run them.
Checklist
Controls implemented.
| Control | How it works |
|---|---|
| Workspace isolation | Row Level Security is enabled on every table that holds customer data. Policies compare each row's organization to the organizations the signed-in user belongs to. |
| Isolation of references | Links between records (an opportunity to its account, for example) use composite foreign keys that include the organization, so a record cannot point at another organization's record. |
| Immutable ownership | The organization on a record cannot be changed after creation; a database trigger rejects the update. |
| Role-based access | Owner, admin and member roles are stored per workspace. Deleting records, managing members and changing roles are restricted by policy and by database functions. |
| File isolation | Call recordings are stored in a private bucket. Storage policies allow access only to objects in a folder named for an organization the user belongs to. |
| Authentication | Sign-up, sign-in, sign-out and password reset are handled by Supabase Auth. Sessions are maintained with cookies, and every application route except the public pages requires a signed-in user. |
| Secrets | The Gemini API key and the service-role key are read on the server only and are never sent to the browser or committed to the repository. |
| Credit and plan integrity | Credit balances and subscriptions cannot be written by users. Credits are deducted by a database function, and plan changes can only be applied by a server-side call authenticated with a secret. |
| Audit trail | Changes to records, AI runs, imports and team changes are written to an activity log visible to workspace members. Members cannot edit or delete entries. |
What this page does not claim
SecureShield AI is a revenue intelligence platform, not a cybersecurity product. It does not perform threat detection, security monitoring or penetration testing, and this page makes no claim of third-party security certification or regulatory compliance. If your review requires specific evidence, contact us and ask.