CVE-2026-110697
How to tell if CVE-2026-110697 (Red Hat Quay robot token theft) is reachable in your environment
Reachable if a self-hosted Red Hat Quay registry runs a build older than the fix for its release line and lets users it does not fully trust create namespaces. CVE-2026-123710, fixed in the same update, also applies wherever the /keys endpoint is reachable by untrusted clients, with no login
On a shared Red Hat Quay registry, any user who can create a namespace can read another team's robot account tokens and use them to push images into that team's repositories. That is CVE-2026-110697, which Red Hat rates 9.9 Critical. Robot tokens are the credentials CI/CD pipelines use to pull and push images, so a stolen token lets an attacker replace an image that production deploys. Red Hat fixed it in Quay 3.9.28, 3.10.28, 3.12.24, 3.14.10, 3.15.9, 3.16.7, 3.17.7 and 3.18.3, and says Quay.io is already fixed.
The same update fixes a second flaw that needs no login at all. Red Hat rates CVE-2026-123710 9.8 Critical and says an unauthenticated attacker can take over registry token signing on any deployment where the /keys endpoint is reachable by untrusted clients.
Public code posted November 13 builds real Quay from Quay's own source, takes another tenant's tokens and overwrites a private image on the vulnerable build. On the fixed build the same steps fail. We read it and did not run it. No exploitation has been reported for either flaw, and neither is on CISA's Known Exploited Vulnerabilities catalog.
Is it reachable for you?
What it is. Red Hat says the robot account listing function builds a database search pattern from the caller's own namespace name without escaping it. A namespace name with an underscore in the right place matches another tenant's namespace, and the listing returns that tenant's robot accounts with their decrypted tokens. Red Hat says no elevated privileges or victim interaction are needed. Any account that can create a namespace is enough, and the public code notes that a user's own namespace counts.
Who is most exposed. Red Hat calls out multi-tenant Quay, meaning one registry shared across teams or by a service provider hosting several customers. It also names OpenShift environments that use Quay as the registry, where stolen robot credentials can put malicious images into CI/CD pipelines.
The companion flaw. CVE-2026-123710 is in Quay's service key server. Red Hat says an unauthenticated client can install its own approved signing key with no expiry, and registry tokens signed with that key are accepted for any repository. That gives read and write access to private images and lets the attacker replace tags. Red Hat says deployments where /keys is reachable by untrusted clients are affected. No public code has been posted for it.
Checks, in three tiers. The first decides exposure. The second and third decide priority. Run these only on registries your organization owns.
- Is the registry on a fixed build? Check the Quay version in the superuser panel or the image tag your deployment runs. The fixed builds are 3.9.28, 3.10.28, 3.12.24, 3.14.10, 3.15.9, 3.16.7, 3.17.7 and 3.18.3. Anything older on the same release line is affected by both flaws.
- Can the /keys endpoint be reached from outside? Test from the internet and from user networks whether the registry's /keys path answers. If it does, CVE-2026-123710 applies, and it needs no account.
- Who can create namespaces? In the registry configuration, check whether ordinary users can create organizations or sign up for accounts, and whether more than one team or customer shares the registry. A shared registry with self-service namespaces is where CVE-2026-110697 matters most. Then list which robot accounts push images that production clusters deploy.
Compensating controls (test them, don't assume them)
Red Hat says updating is the fix and that there is no other complete workaround. These reduce exposure until the update is in, per Red Hat.
- Limit namespace creation. Restrict who can create organizations or namespaces containing an underscore. Red Hat says this reduces exposure to CVE-2026-110697 without fixing it.
- Rotate robot tokens. Rotate the robot tokens of any namespace that collides with a namespace containing an underscore, and watch for pulls or pushes that don't match how those robots are normally used.
- Lock down /keys. For CVE-2026-123710, restrict the /keys endpoint at the proxy or load balancer so only trusted internal components can reach it. Red Hat warns that deleting a rogue key does not last until a fixed build is running, because the same request can install a new one.
What to look for in logs
Neither Red Hat nor anyone else has reported exploitation. These are the checks Red Hat and the public write-up point to.
- Namespaces or user accounts whose names match an existing namespace except for one character replaced by an underscore.
- Robot account listings, pulls or pushes from users outside the namespace that owns the robot.
- For CVE-2026-123710, audit log events for service_key_rotate and service_key_delete with a null expiration_date, and approved service keys your team does not recognize, per Red Hat.
- Image tags in private repositories that changed without a matching pipeline run.
FAQ
Am I exposed to CVE-2026-110697?
You may be if you run self-hosted Red Hat Quay older than the fixed build for its release line and users you don't fully trust can create namespaces on it. Red Hat says multi-tenant registries are most at risk. Quay.io is already fixed.
Is CVE-2026-110697 being exploited?
No exploitation has been reported, and it is not on CISA KEV as of the November 13, 2026 catalog. Public code posted November 13 shows token theft and an image overwrite on a real vulnerable build. We read it and did not run it.
Is there a patch?
Yes. Red Hat fixed it in Quay 3.9.28, 3.10.28, 3.12.24, 3.14.10, 3.15.9, 3.16.7, 3.17.7 and 3.18.3. The same update fixes CVE-2026-123710.
Does it need a login?
CVE-2026-110697 needs an account that can create a namespace, and a user's own namespace is enough. CVE-2026-123710 needs no login, per Red Hat.
We use Quay.io. Do we need to do anything?
Red Hat says Quay.io is already fixed. The checks on this page are for self-hosted Quay.
- Vendor
- Red Hat
- Product
- Red Hat Quay and Project Quay, self-hosted container registry
- Affected versions
- Red Hat Quay 3.9, 3.10, 3.12, 3.14, 3.15, 3.16, 3.17 and 3.18 before the fixed build for each line (3.9.28, 3.10.28, 3.12.24, 3.14.10, 3.15.9, 3.16.7, 3.17.7 and 3.18.3), per Red Hat. Red Hat says Quay.io is already fixed
- Vulnerable component
- Robot account listing API, which builds a database search pattern from the caller's namespace name
- CVSS
- 9.9 Critical (Red Hat, CVSS 3.1). NVD has not added its own score. The companion flaw CVE-2026-123710 is 9.8 Critical (Red Hat)
- CVSS vector
- CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:C/C:H/I:H/A:H (Red Hat)
- CWE
- CWE-89 (Red Hat)
- Exploitation
- Not observed
- CISA KEV added
- Patch
- Red Hat Quay 3.9.28, 3.10.28, 3.12.24, 3.14.10, 3.15.9, 3.16.7, 3.17.7 and 3.18.3 (RHSA-2026:99019 through RHSA-2026:99027). The same update fixes CVE-2026-123710. Upstream fix commit 094cfb0. Quay.io is already fixed
Updates
Nov 16, 2026. First published. Red Hat fixed CVE-2026-110697 and CVE-2026-123710 in Quay 3.9.28, 3.10.28, 3.12.24, 3.14.10, 3.15.9, 3.16.7, 3.17.7 and 3.18.3; public code since November 13 shows cross-tenant token theft on a real vulnerable build; no exploitation reported; neither is on CISA KEV.
Sources
- Red Hat, CVE-2026-110697 (Critical 9.9, published Nov 11, 2026)
- Red Hat, CVE-2026-123710 (Critical 9.8, published Nov 11, 2026)
- Upstream fix commit 094cfb0 (fixes both CVEs)
- NIST NVD: CVE-2026-110697 (Awaiting Analysis, checked Nov 16)
- Public proof of concept, googlesky/CVE-2026-110697 (posted Nov 13, 2026; read only, not run)
- CISA Known Exploited Vulnerabilities catalog (not listed; checked catalog version 2026.11.13)
We publish an advisory when a vulnerability is Critical or on CISA KEV, has one identifiable component to check, and has a public proof of concept or reported exploitation. Advisories are updated as exploitation status changes.