A namespace name became a way to match someone else's account

Red Hat describes a flaw in Quay's robot-account listing function. It uses a namespace name in a SQL LIKE pattern without escaping wildcard characters. An authenticated attacker able to create a specially named namespace could match a different tenant's namespace and obtain its decrypted robot tokens.

Those tokens are not just read access in every deployment. Their permissions may allow an attacker to pull private images, push new ones or replace tags. That turns a registry access-control failure into a potential build and deployment problem. Red Hat rates CVE-2026-110697 critical and says Quay.io has already been fixed.

Known impact is a capability, not a breach count

The advisory explains what an attacker could do. It does not establish that every multi-tenant Quay registry was attacked or that any particular image was changed.

Patch the registry and check the credentials it protects

  1. 1

    Find affected Quay versions. Identify self-hosted and managed Quay deployments. Follow Red Hat's version-specific fixed-release list rather than guessing from the major version.

  2. 2

    Upgrade. Apply the fixed Quay release. Red Hat says restricting namespace creation can reduce exposure temporarily, but it is not a complete workaround.

  3. 3

    Review robot activity. Look for unexpected pulls, pushes or tag changes by robot accounts, especially across tenant boundaries. Rotate tokens where exposure is suspected.

The image digest is more useful than the tag name

If a robot credential could push to a production repository, compare image digests and deployment records with trusted build output. A tag that still has the expected name may point to different content. Keep the investigation tied to the account's actual permissions.