AI Solutions Exchange
Get started
Trust Center

Everything your security, legal, and procurement teams need — in one place.

AI Solutions Exchange is built to be reviewed and approved. Documents, controls, data flows, and a live vendor risk assessment are all published here — no NDA required.

Procurement hubDocument libraryStart vendor risk assessment

Controls

Confirmed controls from the current codebase, partial capabilities under active development, and planned work. We do not list controls we have not built, and we do not claim external assurance we have not obtained.

Authentication & authorization

  • Email and Google authentication via managed provider
    Implemented
  • Default buyer role assigned on account creation
    Implemented
  • Roles stored in a separate table; role checks run through a private-schema security-definer helper
    Implemented
  • Session-gated authenticated routes
    Implemented
  • Protected server functions use authenticated middleware and re-verify caller before privileged access
    Implemented
  • SAML SSO for enterprise workspaces
    Planned
  • MFA enforcement policies at the organization level
    Planned

Protected solution content

Phase 1 / 1b architecture — see the Data flow page for the diagram.

  • Protected content stored in a dedicated table separate from marketplace metadata
    Implemented
  • Direct access revoked from anonymous and ordinary authenticated roles; RLS enabled and forced
    Implemented
  • No buyer-facing RLS policies on the protected content table (deny by default)
    Implemented
  • All writes go through authorized server functions that verify ownership or organization admin role
    Implemented
  • Protected content is not returned from write operations
    Implemented
  • Narrow, auditable reveal-for-editing path for the owner
    Implemented
  • Server-side integrity hash generated on write
    Implemented
  • Version-bound content records; superseded content remains historically traceable
    Implemented
  • Application-level authenticated encryption with key separation and rotation (infrastructure-level encryption at rest applies today)
    Planned

Database security

  • Explicit GRANTs per role on every public-schema table
    Implemented
  • Row-level security enabled on user-scoped tables
    Implemented
  • Forced RLS + deny-by-default on the protected content table
    Implemented
  • Privileged database access restricted to server-side service-role operations
    Implemented
  • Additive migrations; no destructive schema resets as a normal pattern
    Implemented

Protection state machine

  • Protection states: legacy_unclassified, setup_incomplete, protected, suspended
    Implemented
  • Legacy solutions are not automatically labeled protected
    Implemented
  • Server-controlled promotion — a UI selection alone does not grant protected status
    Implemented
  • Readiness check requires active protected content, active setup wizard, and analysis not in a failed state
    Implemented
  • Downgrade if required components become invalid or stale
    Implemented
  • Expanded readiness signals (execution mode availability, publication review workflow)
    Partial

Setup & delivery policy

  • Buyer receives a buyer-safe setup schema; submission stores responses
    Implemented
  • Server validates completeness and applies the delivery policy
    Implemented
  • For protected listings, the wizard response never includes the rendered proprietary prompt
    Implemented
  • Setup submission and solution execution are separate concerns
    Implemented
  • Legacy-distributable delivery only permitted for solutions explicitly classified for it
    Implemented
  • Full protected AI session runtime
    Partial

Public API & extension readiness

  • Buyer-safe manifest endpoint (unified schema)
    Implemented
  • Marketplace listing, search metadata, and setup schema endpoints
    Implemented
  • API-key bearer authentication for protected or mutating endpoints
    Implemented
  • Public manifest never returns protected implementation content
    Implemented
  • Public wizard submission never returns raw prompt for protected solutions
    Implemented
  • Published VS Code, Cursor, Windsurf, and JetBrains adapters
    Planned

API keys

  • Keys issued through authenticated application flows
    Implemented
  • Only SHA-256 hashes stored; raw key shown once at creation
    Implemented
  • Keys are revocable
    Implemented
  • Bearer verification on protected endpoints; possession alone does not grant unrestricted access — endpoints still check entitlement and delivery policy
    Implemented

AI analysis & certification

  • Server-side capability manifest generation
    Implemented
  • Security labels and enterprise risk classification (AI-assisted assessment)
    Implemented
  • Setup-wizard generation and compatibility scoring
    Implemented
  • Certification report (clarity, hallucination risk, completeness, security, maintainability, deployment readiness)
    Implemented
  • Pricing recommendations and marketplace metrics
    Implemented
  • Independent third-party assurance for any of the above (SOC 2, ISO 27001, penetration testing)
    Planned

Auditing

  • Append-only audit_logs on administrative and privileged actions
    Implemented
  • Protected plaintext is not recorded in audit events
    Implemented
  • Organization-scoped audit export
    Partial
  • SIEM / external log-shipping integrations
    Planned

A note on compliance frameworks

The controls above are documented for internal engineering discipline and for enterprise reviewers building their own evidence file. They are not a substitute for an independent audit. AI Solutions Exchange has not completed SOC 2, ISO 27001, or a formal penetration test, and this page does not imply otherwise. When those programs are formally initiated we will publish scope, auditor, and timeline here.