Summary
Capgo contains an improper access control flaw in the public.get_org_members RPC that lets unauthenticated callers list organization members using only a public key and an organization UUID.
This exposure allows attackers to harvest sensitive metadata such as email addresses, user IDs, roles, and pending invitations, especially dangerous in multi-tenant or high-value organizations.
Affected context
The vulnerability occurs in Capgo deployments where:
- The
public.get_org_membersRPC is exposed as part of the backend API surface - Public
sb_publishable_*keys are used to access public-facing endpoints - The endpoint does not enforce strong authentication or authorization tied to organization membership
In this context, any actor that knows (or can guess) an organization UUID can query membership data without being a legitimate member.
Exploit preconditions
- Attack surface: Network-accessible Capgo API endpoint exposing
public.get_org_members - Required access: Possession of a public
sb_publishable_*key, which is intended for unauthenticated/public use - Trigger point: Calling the RPC with a target organization UUID
- Why exploitation is possible: The endpoint relies on insufficient or missing authorization checks, treating requests authenticated only by a public key as authorized to access organization-level membership data
Why this happens
This issue stems from improper access control (CWE-284), where an endpoint that returns sensitive organization membership details is exposed via a public RPC interface without enforcing user-level or organization-level authorization.
The public.get_org_members function effectively treats possession of a public API key and an organization UUID as sufficient to access internal membership metadata, rather than verifying that the caller is an authenticated and authorized member of the organization.
Conceptually, this reflects a design where read operations on security-sensitive resources are misclassified as "public," leading to unintended data disclosure.
Potential impact
- Attackers may enumerate organization members, including email addresses and user IDs
- Attackers could map organization structure, roles, and access patterns
- Increased risk of targeted phishing, social engineering, and credential stuffing attacks
- Exposure of pending invitations can reveal planned collaborations or new hires
- Potential correlation of user identities across organizations and services
General mitigation guidance
Effective mitigation typically involves tightening authorization around organization-scoped data and reducing trust in public keys:
- Require authenticated user sessions for any endpoint that returns organization membership or other sensitive metadata
- Enforce per-organization authorization checks to ensure callers belong to the organization whose members they query
- Separate public, anonymous APIs from organization-internal or user-scoped APIs, with distinct keys and policies
- Minimize data returned by default; expose only non-sensitive fields when broader access is required and gate sensitive fields behind stricter checks
- Apply the principle of least privilege so that clients using public keys cannot access private organizational information
How teams detect issues like this
Detection efforts can focus on both code and runtime behavior:
- Static analysis / code review: Identify RPCs or API routes that:
- Return organization or user metadata
- Are wired to "public" or anonymous authentication contexts
- Lack explicit authorization checks tied to user identity and organization membership
- Security reviews: Examine API contracts for endpoints that take organization identifiers and verify that they enforce membership checks
- Logging and monitoring: Enable detailed logging for calls to organization-related RPCs and flag:
- Repeated access from unknown clients
- High-volume enumeration patterns across different organization UUIDs
- Policy enforcement: Use API gateways or service meshes to apply consistent authentication and authorization policies to organization-scoped endpoints, reducing the chance of misclassified "public" routes

