Summary
Cohere North AI v1.1.5 contains an access-control flaw in its internal user-management API that may allow unauthorized requests to overwrite user information. The issue becomes risky when the endpoint fails to verify that the requester is authorized to modify the targeted user record.
Affected context
The vulnerability occurs in Cohere North AI v1.1.5 at the /api/internal/v1/users/{{USER_ID}} endpoint. It affects deployments where requests reaching this endpoint are not consistently checked against the identity, role, or ownership of the user record identified by USER_ID.
Exploit preconditions
- Attack surface: The internal user-management API endpoint
- Required access: Ability to send requests that reach the affected endpoint; authentication requirements may depend on deployment configuration
- Trigger point: A crafted request targeting a selected user identifier
- Why exploitation is possible: The endpoint may accept user-information updates without adequately enforcing authorization for the targeted record
Why this happens
The underlying issue is an authorization failure in an object-level update operation. The application appears to rely on request reachability or insufficient identity checks instead of confirming that the caller has permission to modify the specific user identified by the request. This behavior maps to CWE-284: Improper Access Control, in which protected operations are not restricted according to the intended security policy.
Potential impact
- An unauthorized user may overwrite account or profile information
- Attackers could disrupt account integrity or user workflows
- Modified user data could support impersonation or social-engineering attempts
- Administrative or operational records could become unreliable
- The vulnerability could contribute to broader compromise if overwritten fields influence authorization or downstream processes
General mitigation guidance
Mitigation generally focuses on enforcing authorization at the point where each user record is accessed or changed.
Common defensive patterns include:
- Apply server-side object-level authorization for every update request
- Derive the target identity from trusted session or token claims where appropriate
- Use deny-by-default policies and explicit role or ownership checks
- Validate that submitted fields are permitted for the caller to change
- Avoid trusting client-supplied identifiers or privilege-related attributes
- Run the service with the minimum permissions needed to perform user-management operations
- Preserve audit records and require additional controls for sensitive account changes
How teams detect issues like this
Organizations can look for this weakness through:
- Static analysis and code review of update handlers for missing authorization checks
- Review of routes that accept user identifiers from URL paths or request bodies
- Tests that attempt cross-user updates using accounts with different roles or ownership
- API gateway and service policies requiring authenticated identity and object-level authorization
- Audit-log monitoring for unusual updates to multiple user records or changes made by unexpected principals
- Automated security testing tools that exercise broken object-level authorization scenarios

