Critical Red Hat Quay flaws expose every private image in a registry
Two critical bugs in Red Hat Quay can let attackers read and overwrite private container images, in one case without an account when anonymous access is on.
Red Hat has disclosed two critical vulnerabilities in Quay, its container image registry, that could give attackers control over private images across an entire registry. Both were found by Red Hat's Dave O'Connor and affect all Quay 3 and Quay 5 versions.
The two flaws
CVE-2026-123710, rated 9.8, lets an attacker add their own signing key to Quay's list of trusted keys, with no expiry date. Tokens signed with that key grant read and overwrite access to private images in every repository. Because anonymous access is on by default, an attacker may not even need an account. The planted key also survives restarts and password resets.
CVE-2026-110697, rated 9.9, comes from unescaped wildcards in a database search. A user who can create a namespace can steal other tenants' robot account tokens and then read, push or overwrite their images.
Why it matters
Container registries feed production systems. An attacker who can overwrite images could slip malicious code into software that gets deployed automatically. Red Hat has released fixed builds for several Quay 3 branches, and the hosted Quay.io service is already fixed. There are no reports of exploitation so far.
Who's affected
Organizations running self-hosted Red Hat Quay 3 or Quay 5 registries, especially those with anonymous access enabled or many tenants.
What to do
Install the fixed Quay release. Until then, block the /keys endpoint at the proxy except for trusted internal components, search audit logs for service key events with no expiration date, remove unknown keys, and rotate robot tokens.