This CVE exposes a fundamental authorization gap in Open Cluster Management (OCM) and Advanced Cluster Management (ACM): the system treats cluster targeting as a separate concern from application creation, when in a multi-tenant context they should be inseparable. A user with permission to create Application custom resources on the hub cluster can inject the ocm-managed-cluster annotation to direct propagation to any managed cluster in the federation—regardless of whether they have any business targeting that cluster. The propagation controller faithfully carries out the targeting instruction without verifying that the creator's RBAC profile authorizes cluster selection.
This isn't a subtle implementation bug. It's an architectural flaw where RBAC governs where you can create objects, but the annotation handler governs where those objects propagate. A tenant with hub-level Application CREATE can pivot through the hub's propagation controller to land code on all managed clusters simultaneously—a multiplicative blast radius that makes the hub a single point of failure for the entire federation.
The fix requires annotation-level validation tied to cluster targeting permissions. But the practical remediation burden extends far beyond patching. Platform teams must audit every existing Application CR in their hub for malicious or unintentional cluster targeting annotations—the vulnerability doesn't close at patch time, it closes at the next reconciliation cycle of all affected Applications. This is compounded remediation debt: the CVE patch is a two-hour change, but the customer audit is a six-month conversation.
The deeper concern: this pattern—annotations controlling security-sensitive cluster targeting without explicit RBAC scoping—likely exists in other multi-cluster management tools (Rancher, Anthos, Tanzu). The same vulnerability class was documented in Kubernetes admission controller bypasses circa 2018-2019 but never propagated into OCM/ACM's architecture reviews. Organizations running multi-tenant federations should audit their tooling for this exact pattern: RBAC-governed object creation combined with annotation-based propagation control.