CVE-2026-100652

HIGH 8.2

Published 2026-09-26 · Last modified 2026-09-28

vLLM versions 0.22.0 through 0.23.0 fail to validate stop_token_ids against vocabulary bounds in Rust HTTP and gRPC frontends, allowing out-of-vocabulary token IDs to reach MinTokensLogitsProcessor. Attackers can submit requests with min_tokens greater than zero and out-of-vocabulary stop_token_ids to trigger CUDA tensor indexing failures that leave EngineCore in a fatal state requiring service restart.

ELEVATED IMPACT

Severe if exploited (CVSS 8.2), but no known exploitation and low modeled probability. Patch on a normal cadence.

Exploitation likelihood

0.4%chance of exploitation in 30 days · 36th percentile

○ In CISA KEV ○ Catalogued PoC

Impact if exploited

8.2CVSS 4.0 · HIGH

ConfidentialityNone
IntegrityNone
AvailabilityHigh

What an attacker needs

✓Access: Reachable over the network — no local access needed
✓Privileges: No account or privileges required
✓User interaction: No user interaction needed
⚠Complexity: Needs a race window or specific setup
⚠Requirements: Specific conditions must be present

✓ lowers the bar for an attacker · ⚠ raises it

Proof of concept

No public PoC is currently catalogued.

Search for new PoCs

Search GitHub now — results may be empty or unverified.

Tracked sources are checked daily for recent CVEs, including PoCs published after the original disclosure. A later release appears after a tracked source lists it and a refresh succeeds.

Data snapshot: 2026-10-26T13:52:32.101Z

Affected

Vendors Vllm Project Vllm

Products Vllm

Weakness (CWE)

CWE-20: Improper Input Validation

CVSS vector

CVSS:4.0/AV:N/AC:H/AT:P/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N

References and research data (JSON)

Packages and version ranges

Each advisory describes its own package and affected ranges. A boundary applies to that range; it is not a universal upgrade recommendation.

vllm PyPI

pkg:pypi/vllm
ECOSYSTEM range
- Introduced version (included): 0.22.0
- Fixed boundary (excluded): 0.24.0
3 explicitly listed affected versions
0.22.0, 0.22.1, 0.23.0
Additional upstream package constraints
```
{
  "ecosystem": {},
  "database": {
    "source": "https://github.com/pypa/advisory-database/blob/main/vulns/vllm/PYSEC-2026-4185.yaml"
  }
}
```
PyPA · PYSEC-2026-4185 · Updated 2026-10-07T10:00:03.378975746Z

Vendor product assessments

Showing 32 of 344 vendor product statements; 312 omitted. Read the full vendor advisory.

registry.redhat.io/rhoai/odh-kserve-agent-rhel9@sha256:18eb52b8f8ed403671396e49370216a8a553d79dfe33efb3cae79284eb40643a_ppc64le as a component of Red Hat OpenShift AI 3.4 · fixed

Vendor assessment for this product only.

pkg:oci/odh-kserve-agent-rhel9@sha256%3A18eb52b8f8ed403671396e49370216a8a553d79dfe33efb3cae79284eb40643a?arch=ppc64le&repository_url=registry.redhat.io/rhoai/odh-kserve-agent-rhel9&tag=1791468381
- vendor fix: For Red Hat OpenShift AI 3.4.5 see the following documentation, which will be updated shortly for this release, for important instructions on how to upgrade your cluster and fully apply this errata update: https://docs.redhat.com/en/documentation/red_hat_openshift_ai/ Vendor guidance
- workaround: Restrict network access to vLLM HTTP and gRPC service endpoints to trusted networks and authenticated clients only. In host-based environments, restrict the serving port (typically TCP port 8000) using firewalld: ``` firewall-cmd --permanent --zone=trusted --add-source=<trusted_ip_or_subnet> firewall-cmd --permanent --zone=public --remove-port=8000/tcp firewall-cmd --reload ``` In Red Hat OpenShift AI deployments, apply a Kubernetes NetworkPolicy to the model-serving namespace to ensure only authorized ingress controllers or trusted application pods can reach the inference containers. Warning: Reloading firewall rules with `firewall-cmd --reload` applies changes immediately and may interrupt active network connections. Caveats: Network-level controls limit exposure to external untrusted callers, but they do not protect against crafted requests submitted by authorized users or internal services that consume the inference API.
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
registry.redhat.io/rhoai/odh-kserve-agent-rhel9@sha256:636ba382740d2c14ace1c03d8474e460dc78929237e1efae7bd2da58605fb9f8_s390x as a component of Red Hat OpenShift AI 3.4 · fixed

Vendor assessment for this product only.

pkg:oci/odh-kserve-agent-rhel9@sha256%3A636ba382740d2c14ace1c03d8474e460dc78929237e1efae7bd2da58605fb9f8?arch=s390x&repository_url=registry.redhat.io/rhoai/odh-kserve-agent-rhel9&tag=1791468381
- vendor fix: For Red Hat OpenShift AI 3.4.5 see the following documentation, which will be updated shortly for this release, for important instructions on how to upgrade your cluster and fully apply this errata update: https://docs.redhat.com/en/documentation/red_hat_openshift_ai/ Vendor guidance
- workaround: Restrict network access to vLLM HTTP and gRPC service endpoints to trusted networks and authenticated clients only. In host-based environments, restrict the serving port (typically TCP port 8000) using firewalld: ``` firewall-cmd --permanent --zone=trusted --add-source=<trusted_ip_or_subnet> firewall-cmd --permanent --zone=public --remove-port=8000/tcp firewall-cmd --reload ``` In Red Hat OpenShift AI deployments, apply a Kubernetes NetworkPolicy to the model-serving namespace to ensure only authorized ingress controllers or trusted application pods can reach the inference containers. Warning: Reloading firewall rules with `firewall-cmd --reload` applies changes immediately and may interrupt active network connections. Caveats: Network-level controls limit exposure to external untrusted callers, but they do not protect against crafted requests submitted by authorized users or internal services that consume the inference API.
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
registry.redhat.io/rhoai/odh-kserve-agent-rhel9@sha256:e7063293523536dcb008ff4e9479d05b40a9caa9b0f189318ff1c8362d08602a_amd64 as a component of Red Hat OpenShift AI 3.4 · fixed

Vendor assessment for this product only.

pkg:oci/odh-kserve-agent-rhel9@sha256%3Ae7063293523536dcb008ff4e9479d05b40a9caa9b0f189318ff1c8362d08602a?arch=amd64&repository_url=registry.redhat.io/rhoai/odh-kserve-agent-rhel9&tag=1791468381
- vendor fix: For Red Hat OpenShift AI 3.4.5 see the following documentation, which will be updated shortly for this release, for important instructions on how to upgrade your cluster and fully apply this errata update: https://docs.redhat.com/en/documentation/red_hat_openshift_ai/ Vendor guidance
- workaround: Restrict network access to vLLM HTTP and gRPC service endpoints to trusted networks and authenticated clients only. In host-based environments, restrict the serving port (typically TCP port 8000) using firewalld: ``` firewall-cmd --permanent --zone=trusted --add-source=<trusted_ip_or_subnet> firewall-cmd --permanent --zone=public --remove-port=8000/tcp firewall-cmd --reload ``` In Red Hat OpenShift AI deployments, apply a Kubernetes NetworkPolicy to the model-serving namespace to ensure only authorized ingress controllers or trusted application pods can reach the inference containers. Warning: Reloading firewall rules with `firewall-cmd --reload` applies changes immediately and may interrupt active network connections. Caveats: Network-level controls limit exposure to external untrusted callers, but they do not protect against crafted requests submitted by authorized users or internal services that consume the inference API.
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
registry.redhat.io/rhoai/odh-kserve-controller-rhel9@sha256:51997b1a7e5e2f3c0bd76b7d23fda33902843addafcbba6333b27d5c72beb5c4_amd64 as a component of Red Hat OpenShift AI 3.4 · fixed

Vendor assessment for this product only.

pkg:oci/odh-kserve-controller-rhel9@sha256%3A51997b1a7e5e2f3c0bd76b7d23fda33902843addafcbba6333b27d5c72beb5c4?arch=amd64&repository_url=registry.redhat.io/rhoai/odh-kserve-controller-rhel9&tag=1791468368
- vendor fix: For Red Hat OpenShift AI 3.4.5 see the following documentation, which will be updated shortly for this release, for important instructions on how to upgrade your cluster and fully apply this errata update: https://docs.redhat.com/en/documentation/red_hat_openshift_ai/ Vendor guidance
- workaround: Restrict network access to vLLM HTTP and gRPC service endpoints to trusted networks and authenticated clients only. In host-based environments, restrict the serving port (typically TCP port 8000) using firewalld: ``` firewall-cmd --permanent --zone=trusted --add-source=<trusted_ip_or_subnet> firewall-cmd --permanent --zone=public --remove-port=8000/tcp firewall-cmd --reload ``` In Red Hat OpenShift AI deployments, apply a Kubernetes NetworkPolicy to the model-serving namespace to ensure only authorized ingress controllers or trusted application pods can reach the inference containers. Warning: Reloading firewall rules with `firewall-cmd --reload` applies changes immediately and may interrupt active network connections. Caveats: Network-level controls limit exposure to external untrusted callers, but they do not protect against crafted requests submitted by authorized users or internal services that consume the inference API.
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
registry.redhat.io/rhoai/odh-kserve-controller-rhel9@sha256:a2967b5e30b4b8c6299abde8dc0fcab8ceb15581aff94dda4f02ac6e6aa05cb7_s390x as a component of Red Hat OpenShift AI 3.4 · fixed

Vendor assessment for this product only.

pkg:oci/odh-kserve-controller-rhel9@sha256%3Aa2967b5e30b4b8c6299abde8dc0fcab8ceb15581aff94dda4f02ac6e6aa05cb7?arch=s390x&repository_url=registry.redhat.io/rhoai/odh-kserve-controller-rhel9&tag=1791468368
- vendor fix: For Red Hat OpenShift AI 3.4.5 see the following documentation, which will be updated shortly for this release, for important instructions on how to upgrade your cluster and fully apply this errata update: https://docs.redhat.com/en/documentation/red_hat_openshift_ai/ Vendor guidance
- workaround: Restrict network access to vLLM HTTP and gRPC service endpoints to trusted networks and authenticated clients only. In host-based environments, restrict the serving port (typically TCP port 8000) using firewalld: ``` firewall-cmd --permanent --zone=trusted --add-source=<trusted_ip_or_subnet> firewall-cmd --permanent --zone=public --remove-port=8000/tcp firewall-cmd --reload ``` In Red Hat OpenShift AI deployments, apply a Kubernetes NetworkPolicy to the model-serving namespace to ensure only authorized ingress controllers or trusted application pods can reach the inference containers. Warning: Reloading firewall rules with `firewall-cmd --reload` applies changes immediately and may interrupt active network connections. Caveats: Network-level controls limit exposure to external untrusted callers, but they do not protect against crafted requests submitted by authorized users or internal services that consume the inference API.
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
registry.redhat.io/rhoai/odh-kserve-router-rhel9@sha256:24cf45f8cc50e378b6a7ecb0f42568f6426242af8b2751d60350d04b3817ccd4_s390x as a component of Red Hat OpenShift AI 3.4 · fixed

Vendor assessment for this product only.

pkg:oci/odh-kserve-router-rhel9@sha256%3A24cf45f8cc50e378b6a7ecb0f42568f6426242af8b2751d60350d04b3817ccd4?arch=s390x&repository_url=registry.redhat.io/rhoai/odh-kserve-router-rhel9&tag=1791468401
- vendor fix: For Red Hat OpenShift AI 3.4.5 see the following documentation, which will be updated shortly for this release, for important instructions on how to upgrade your cluster and fully apply this errata update: https://docs.redhat.com/en/documentation/red_hat_openshift_ai/ Vendor guidance
- workaround: Restrict network access to vLLM HTTP and gRPC service endpoints to trusted networks and authenticated clients only. In host-based environments, restrict the serving port (typically TCP port 8000) using firewalld: ``` firewall-cmd --permanent --zone=trusted --add-source=<trusted_ip_or_subnet> firewall-cmd --permanent --zone=public --remove-port=8000/tcp firewall-cmd --reload ``` In Red Hat OpenShift AI deployments, apply a Kubernetes NetworkPolicy to the model-serving namespace to ensure only authorized ingress controllers or trusted application pods can reach the inference containers. Warning: Reloading firewall rules with `firewall-cmd --reload` applies changes immediately and may interrupt active network connections. Caveats: Network-level controls limit exposure to external untrusted callers, but they do not protect against crafted requests submitted by authorized users or internal services that consume the inference API.
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
registry.redhat.io/rhoai/odh-kserve-router-rhel9@sha256:3d602b9c38076730a0972ed143ff886d2018ef9ffbcc9ca339ad8dade8bfb05f_ppc64le as a component of Red Hat OpenShift AI 3.4 · fixed

Vendor assessment for this product only.

pkg:oci/odh-kserve-router-rhel9@sha256%3A3d602b9c38076730a0972ed143ff886d2018ef9ffbcc9ca339ad8dade8bfb05f?arch=ppc64le&repository_url=registry.redhat.io/rhoai/odh-kserve-router-rhel9&tag=1791468401
- vendor fix: For Red Hat OpenShift AI 3.4.5 see the following documentation, which will be updated shortly for this release, for important instructions on how to upgrade your cluster and fully apply this errata update: https://docs.redhat.com/en/documentation/red_hat_openshift_ai/ Vendor guidance
- workaround: Restrict network access to vLLM HTTP and gRPC service endpoints to trusted networks and authenticated clients only. In host-based environments, restrict the serving port (typically TCP port 8000) using firewalld: ``` firewall-cmd --permanent --zone=trusted --add-source=<trusted_ip_or_subnet> firewall-cmd --permanent --zone=public --remove-port=8000/tcp firewall-cmd --reload ``` In Red Hat OpenShift AI deployments, apply a Kubernetes NetworkPolicy to the model-serving namespace to ensure only authorized ingress controllers or trusted application pods can reach the inference containers. Warning: Reloading firewall rules with `firewall-cmd --reload` applies changes immediately and may interrupt active network connections. Caveats: Network-level controls limit exposure to external untrusted callers, but they do not protect against crafted requests submitted by authorized users or internal services that consume the inference API.
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
registry.redhat.io/rhoai/odh-kserve-router-rhel9@sha256:96ddd9921ce58a1df83c1a89a2046651feaeb2db91e8f48740335ac7096a3457_arm64 as a component of Red Hat OpenShift AI 3.4 · fixed

Vendor assessment for this product only.

pkg:oci/odh-kserve-router-rhel9@sha256%3A96ddd9921ce58a1df83c1a89a2046651feaeb2db91e8f48740335ac7096a3457?arch=arm64&repository_url=registry.redhat.io/rhoai/odh-kserve-router-rhel9&tag=1791468401
- vendor fix: For Red Hat OpenShift AI 3.4.5 see the following documentation, which will be updated shortly for this release, for important instructions on how to upgrade your cluster and fully apply this errata update: https://docs.redhat.com/en/documentation/red_hat_openshift_ai/ Vendor guidance
- workaround: Restrict network access to vLLM HTTP and gRPC service endpoints to trusted networks and authenticated clients only. In host-based environments, restrict the serving port (typically TCP port 8000) using firewalld: ``` firewall-cmd --permanent --zone=trusted --add-source=<trusted_ip_or_subnet> firewall-cmd --permanent --zone=public --remove-port=8000/tcp firewall-cmd --reload ``` In Red Hat OpenShift AI deployments, apply a Kubernetes NetworkPolicy to the model-serving namespace to ensure only authorized ingress controllers or trusted application pods can reach the inference containers. Warning: Reloading firewall rules with `firewall-cmd --reload` applies changes immediately and may interrupt active network connections. Caveats: Network-level controls limit exposure to external untrusted callers, but they do not protect against crafted requests submitted by authorized users or internal services that consume the inference API.
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
registry.redhat.io/rhoai/odh-kserve-storage-initializer-rhel9@sha256:006fea51fee2672e37e526aa1d0fdd7e6f6f2dcf428fac12a2cc18b5b8ee8e02_s390x as a component of Red Hat OpenShift AI 3.4 · fixed

Vendor assessment for this product only.

pkg:oci/odh-kserve-storage-initializer-rhel9@sha256%3A006fea51fee2672e37e526aa1d0fdd7e6f6f2dcf428fac12a2cc18b5b8ee8e02?arch=s390x&repository_url=registry.redhat.io/rhoai/odh-kserve-storage-initializer-rhel9&tag=1790958549
- vendor fix: For Red Hat OpenShift AI 3.4.5 see the following documentation, which will be updated shortly for this release, for important instructions on how to upgrade your cluster and fully apply this errata update: https://docs.redhat.com/en/documentation/red_hat_openshift_ai/ Vendor guidance
- workaround: Restrict network access to vLLM HTTP and gRPC service endpoints to trusted networks and authenticated clients only. In host-based environments, restrict the serving port (typically TCP port 8000) using firewalld: ``` firewall-cmd --permanent --zone=trusted --add-source=<trusted_ip_or_subnet> firewall-cmd --permanent --zone=public --remove-port=8000/tcp firewall-cmd --reload ``` In Red Hat OpenShift AI deployments, apply a Kubernetes NetworkPolicy to the model-serving namespace to ensure only authorized ingress controllers or trusted application pods can reach the inference containers. Warning: Reloading firewall rules with `firewall-cmd --reload` applies changes immediately and may interrupt active network connections. Caveats: Network-level controls limit exposure to external untrusted callers, but they do not protect against crafted requests submitted by authorized users or internal services that consume the inference API.
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
registry.redhat.io/rhoai/odh-kserve-storage-initializer-rhel9@sha256:35a57dd79362703b69e563f84d18d2ec31359628a246e3c36b52ece728aa9f21_arm64 as a component of Red Hat OpenShift AI 3.4 · fixed

Vendor assessment for this product only.

pkg:oci/odh-kserve-storage-initializer-rhel9@sha256%3A35a57dd79362703b69e563f84d18d2ec31359628a246e3c36b52ece728aa9f21?arch=arm64&repository_url=registry.redhat.io/rhoai/odh-kserve-storage-initializer-rhel9&tag=1790958549
- vendor fix: For Red Hat OpenShift AI 3.4.5 see the following documentation, which will be updated shortly for this release, for important instructions on how to upgrade your cluster and fully apply this errata update: https://docs.redhat.com/en/documentation/red_hat_openshift_ai/ Vendor guidance
- workaround: Restrict network access to vLLM HTTP and gRPC service endpoints to trusted networks and authenticated clients only. In host-based environments, restrict the serving port (typically TCP port 8000) using firewalld: ``` firewall-cmd --permanent --zone=trusted --add-source=<trusted_ip_or_subnet> firewall-cmd --permanent --zone=public --remove-port=8000/tcp firewall-cmd --reload ``` In Red Hat OpenShift AI deployments, apply a Kubernetes NetworkPolicy to the model-serving namespace to ensure only authorized ingress controllers or trusted application pods can reach the inference containers. Warning: Reloading firewall rules with `firewall-cmd --reload` applies changes immediately and may interrupt active network connections. Caveats: Network-level controls limit exposure to external untrusted callers, but they do not protect against crafted requests submitted by authorized users or internal services that consume the inference API.
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
registry.redhat.io/rhoai/odh-kserve-storage-initializer-rhel9@sha256:856c7748eb45dbd190fb16fe90a09546ac8febf2f83e41d8068e8ac759442018_amd64 as a component of Red Hat OpenShift AI 3.4 · fixed

Vendor assessment for this product only.

pkg:oci/odh-kserve-storage-initializer-rhel9@sha256%3A856c7748eb45dbd190fb16fe90a09546ac8febf2f83e41d8068e8ac759442018?arch=amd64&repository_url=registry.redhat.io/rhoai/odh-kserve-storage-initializer-rhel9&tag=1790958549
- vendor fix: For Red Hat OpenShift AI 3.4.5 see the following documentation, which will be updated shortly for this release, for important instructions on how to upgrade your cluster and fully apply this errata update: https://docs.redhat.com/en/documentation/red_hat_openshift_ai/ Vendor guidance
- workaround: Restrict network access to vLLM HTTP and gRPC service endpoints to trusted networks and authenticated clients only. In host-based environments, restrict the serving port (typically TCP port 8000) using firewalld: ``` firewall-cmd --permanent --zone=trusted --add-source=<trusted_ip_or_subnet> firewall-cmd --permanent --zone=public --remove-port=8000/tcp firewall-cmd --reload ``` In Red Hat OpenShift AI deployments, apply a Kubernetes NetworkPolicy to the model-serving namespace to ensure only authorized ingress controllers or trusted application pods can reach the inference containers. Warning: Reloading firewall rules with `firewall-cmd --reload` applies changes immediately and may interrupt active network connections. Caveats: Network-level controls limit exposure to external untrusted callers, but they do not protect against crafted requests submitted by authorized users or internal services that consume the inference API.
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
rhaii/vllm-cpu-rhel9 as a component of Red Hat AI Inference Server · known affected

Vendor assessment for this product only.

pkg:oci/vllm-cpu-rhel9@sha256:81c5994dc20321f8bd3969099eade78ef127cbd79f2d0ccea101c773e4c7fa4d?repository_url=registry.redhat.io/rhaii/vllm-cpu-rhel9
- workaround: Restrict network access to vLLM HTTP and gRPC service endpoints to trusted networks and authenticated clients only. In host-based environments, restrict the serving port (typically TCP port 8000) using firewalld: ``` firewall-cmd --permanent --zone=trusted --add-source=<trusted_ip_or_subnet> firewall-cmd --permanent --zone=public --remove-port=8000/tcp firewall-cmd --reload ``` In Red Hat OpenShift AI deployments, apply a Kubernetes NetworkPolicy to the model-serving namespace to ensure only authorized ingress controllers or trusted application pods can reach the inference containers. Warning: Reloading firewall rules with `firewall-cmd --reload` applies changes immediately and may interrupt active network connections. Caveats: Network-level controls limit exposure to external untrusted callers, but they do not protect against crafted requests submitted by authorized users or internal services that consume the inference API.
- none available: Affected
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
rhaii/vllm-gaudi-rhel9 as a component of Red Hat AI Inference Server · known affected

Vendor assessment for this product only.

pkg:oci/vllm-gaudi-rhel9@sha256:ce1e4fd82dd37d299ce8c64a237971d00faf602f06868c986e81aefa04c283ef?repository_url=registry.redhat.io/rhaii/vllm-gaudi-rhel9
- workaround: Restrict network access to vLLM HTTP and gRPC service endpoints to trusted networks and authenticated clients only. In host-based environments, restrict the serving port (typically TCP port 8000) using firewalld: ``` firewall-cmd --permanent --zone=trusted --add-source=<trusted_ip_or_subnet> firewall-cmd --permanent --zone=public --remove-port=8000/tcp firewall-cmd --reload ``` In Red Hat OpenShift AI deployments, apply a Kubernetes NetworkPolicy to the model-serving namespace to ensure only authorized ingress controllers or trusted application pods can reach the inference containers. Warning: Reloading firewall rules with `firewall-cmd --reload` applies changes immediately and may interrupt active network connections. Caveats: Network-level controls limit exposure to external untrusted callers, but they do not protect against crafted requests submitted by authorized users or internal services that consume the inference API.
- no fix planned: Will not fix
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
rhaii/vllm-rocm-rhel9 as a component of Red Hat AI Inference Server · known affected

Vendor assessment for this product only.

pkg:oci/vllm-rocm-rhel9@sha256:1a30359d6ca5d0891e403b9b0a23c0ae1ad6e30a2429b9b42e246b874fe8859a?repository_url=registry.redhat.io/rhaii/vllm-rocm-rhel9
- workaround: Restrict network access to vLLM HTTP and gRPC service endpoints to trusted networks and authenticated clients only. In host-based environments, restrict the serving port (typically TCP port 8000) using firewalld: ``` firewall-cmd --permanent --zone=trusted --add-source=<trusted_ip_or_subnet> firewall-cmd --permanent --zone=public --remove-port=8000/tcp firewall-cmd --reload ``` In Red Hat OpenShift AI deployments, apply a Kubernetes NetworkPolicy to the model-serving namespace to ensure only authorized ingress controllers or trusted application pods can reach the inference containers. Warning: Reloading firewall rules with `firewall-cmd --reload` applies changes immediately and may interrupt active network connections. Caveats: Network-level controls limit exposure to external untrusted callers, but they do not protect against crafted requests submitted by authorized users or internal services that consume the inference API.
- none available: Affected
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
rhaii/vllm-tpu-rhel9 as a component of Red Hat AI Inference Server · known affected

Vendor assessment for this product only.

pkg:oci/vllm-tpu-rhel9@sha256:0c2ac747d67dc0e80162ff05867dd702e474b2217659f19f28b8e3a0c03a942a?repository_url=registry.redhat.io/rhaii/vllm-tpu-rhel9
- workaround: Restrict network access to vLLM HTTP and gRPC service endpoints to trusted networks and authenticated clients only. In host-based environments, restrict the serving port (typically TCP port 8000) using firewalld: ``` firewall-cmd --permanent --zone=trusted --add-source=<trusted_ip_or_subnet> firewall-cmd --permanent --zone=public --remove-port=8000/tcp firewall-cmd --reload ``` In Red Hat OpenShift AI deployments, apply a Kubernetes NetworkPolicy to the model-serving namespace to ensure only authorized ingress controllers or trusted application pods can reach the inference containers. Warning: Reloading firewall rules with `firewall-cmd --reload` applies changes immediately and may interrupt active network connections. Caveats: Network-level controls limit exposure to external untrusted callers, but they do not protect against crafted requests submitted by authorized users or internal services that consume the inference API.
- no fix planned: Will not fix
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
rhaiis/vllm-cuda-rhel9 as a component of Red Hat AI Inference Server · known affected

Vendor assessment for this product only.

pkg:oci/vllm-cuda-rhel9@sha256:73bbc4560132c4aaa9bbd608ce9560eeae7b30bcc8f014da460afa5622139372?repository_url=registry.redhat.io/rhaiis/vllm-cuda-rhel9
- workaround: Restrict network access to vLLM HTTP and gRPC service endpoints to trusted networks and authenticated clients only. In host-based environments, restrict the serving port (typically TCP port 8000) using firewalld: ``` firewall-cmd --permanent --zone=trusted --add-source=<trusted_ip_or_subnet> firewall-cmd --permanent --zone=public --remove-port=8000/tcp firewall-cmd --reload ``` In Red Hat OpenShift AI deployments, apply a Kubernetes NetworkPolicy to the model-serving namespace to ensure only authorized ingress controllers or trusted application pods can reach the inference containers. Warning: Reloading firewall rules with `firewall-cmd --reload` applies changes immediately and may interrupt active network connections. Caveats: Network-level controls limit exposure to external untrusted callers, but they do not protect against crafted requests submitted by authorized users or internal services that consume the inference API.
- none available: Affected
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
rhaiis/vllm-rocm-rhel9 as a component of Red Hat AI Inference Server · known affected

Vendor assessment for this product only.

pkg:oci/vllm-rocm-rhel9@sha256:37b64a8df90756aae0ee96d2f07765adbce5ed5beb2414ac3845806ad2219cfc?repository_url=registry.redhat.io/rhaiis/vllm-rocm-rhel9
- workaround: Restrict network access to vLLM HTTP and gRPC service endpoints to trusted networks and authenticated clients only. In host-based environments, restrict the serving port (typically TCP port 8000) using firewalld: ``` firewall-cmd --permanent --zone=trusted --add-source=<trusted_ip_or_subnet> firewall-cmd --permanent --zone=public --remove-port=8000/tcp firewall-cmd --reload ``` In Red Hat OpenShift AI deployments, apply a Kubernetes NetworkPolicy to the model-serving namespace to ensure only authorized ingress controllers or trusted application pods can reach the inference containers. Warning: Reloading firewall rules with `firewall-cmd --reload` applies changes immediately and may interrupt active network connections. Caveats: Network-level controls limit exposure to external untrusted callers, but they do not protect against crafted requests submitted by authorized users or internal services that consume the inference API.
- none available: Affected
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
rhelai3/bootc-aws-cuda-rhel9 as a component of Red Hat Enterprise Linux AI (RHEL AI) 3 · known affected

Vendor assessment for this product only.

pkg:oci/bootc-aws-cuda-rhel9@sha256:a754d4d1126e98f203249414b3831182bb15ee6b3643452f167849defc642e4b?repository_url=registry.redhat.io/rhelai3/bootc-aws-cuda-rhel9
- workaround: Restrict network access to vLLM HTTP and gRPC service endpoints to trusted networks and authenticated clients only. In host-based environments, restrict the serving port (typically TCP port 8000) using firewalld: ``` firewall-cmd --permanent --zone=trusted --add-source=<trusted_ip_or_subnet> firewall-cmd --permanent --zone=public --remove-port=8000/tcp firewall-cmd --reload ``` In Red Hat OpenShift AI deployments, apply a Kubernetes NetworkPolicy to the model-serving namespace to ensure only authorized ingress controllers or trusted application pods can reach the inference containers. Warning: Reloading firewall rules with `firewall-cmd --reload` applies changes immediately and may interrupt active network connections. Caveats: Network-level controls limit exposure to external untrusted callers, but they do not protect against crafted requests submitted by authorized users or internal services that consume the inference API.
- none available: Affected
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
rhelai3/bootc-azure-rocm-rhel9 as a component of Red Hat Enterprise Linux AI (RHEL AI) 3 · known affected

Vendor assessment for this product only.

pkg:oci/bootc-azure-rocm-rhel9@sha256:8ea1233b1fb6b82d0aa69e9e4c47d2df242001280c11a61584066c27c5bec3a5?repository_url=registry.redhat.io/rhelai3/bootc-azure-rocm-rhel9
- workaround: Restrict network access to vLLM HTTP and gRPC service endpoints to trusted networks and authenticated clients only. In host-based environments, restrict the serving port (typically TCP port 8000) using firewalld: ``` firewall-cmd --permanent --zone=trusted --add-source=<trusted_ip_or_subnet> firewall-cmd --permanent --zone=public --remove-port=8000/tcp firewall-cmd --reload ``` In Red Hat OpenShift AI deployments, apply a Kubernetes NetworkPolicy to the model-serving namespace to ensure only authorized ingress controllers or trusted application pods can reach the inference containers. Warning: Reloading firewall rules with `firewall-cmd --reload` applies changes immediately and may interrupt active network connections. Caveats: Network-level controls limit exposure to external untrusted callers, but they do not protect against crafted requests submitted by authorized users or internal services that consume the inference API.
- none available: Affected
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
rhelai3/bootc-gaudi-rhel9 as a component of Red Hat Enterprise Linux AI (RHEL AI) 3 · known affected

Vendor assessment for this product only.

pkg:oci/bootc-gaudi-rhel9@sha256:e1057b59ca1be9501332abf97230855e85c7df858356be0da9dce230ea7e98e1?repository_url=registry.redhat.io/rhelai3/bootc-gaudi-rhel9
- workaround: Restrict network access to vLLM HTTP and gRPC service endpoints to trusted networks and authenticated clients only. In host-based environments, restrict the serving port (typically TCP port 8000) using firewalld: ``` firewall-cmd --permanent --zone=trusted --add-source=<trusted_ip_or_subnet> firewall-cmd --permanent --zone=public --remove-port=8000/tcp firewall-cmd --reload ``` In Red Hat OpenShift AI deployments, apply a Kubernetes NetworkPolicy to the model-serving namespace to ensure only authorized ingress controllers or trusted application pods can reach the inference containers. Warning: Reloading firewall rules with `firewall-cmd --reload` applies changes immediately and may interrupt active network connections. Caveats: Network-level controls limit exposure to external untrusted callers, but they do not protect against crafted requests submitted by authorized users or internal services that consume the inference API.
- no fix planned: Will not fix
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
rhelai3/bootc-rocm-rhel9 as a component of Red Hat Enterprise Linux AI (RHEL AI) 3 · known affected

Vendor assessment for this product only.

pkg:oci/bootc-rocm-rhel9@sha256:38ce057f97034bbb8eaea574d41cc7eebaeaa9265a4cdd966476dd3fac129c3c?repository_url=registry.redhat.io/rhelai3/bootc-rocm-rhel9
- workaround: Restrict network access to vLLM HTTP and gRPC service endpoints to trusted networks and authenticated clients only. In host-based environments, restrict the serving port (typically TCP port 8000) using firewalld: ``` firewall-cmd --permanent --zone=trusted --add-source=<trusted_ip_or_subnet> firewall-cmd --permanent --zone=public --remove-port=8000/tcp firewall-cmd --reload ``` In Red Hat OpenShift AI deployments, apply a Kubernetes NetworkPolicy to the model-serving namespace to ensure only authorized ingress controllers or trusted application pods can reach the inference containers. Warning: Reloading firewall rules with `firewall-cmd --reload` applies changes immediately and may interrupt active network connections. Caveats: Network-level controls limit exposure to external untrusted callers, but they do not protect against crafted requests submitted by authorized users or internal services that consume the inference API.
- none available: Affected
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
rhoai/odh-vllm-gaudi-rhel9 as a component of Red Hat OpenShift AI (RHOAI) · known affected

Vendor assessment for this product only.

pkg:oci/odh-vllm-gaudi-rhel9@sha256:87eec3862eb6ddef445f89dcf6fd89add7ef202ea7c563b1ee7347537fb3e250?repository_url=registry.redhat.io/rhoai/odh-vllm-gaudi-rhel9
- workaround: Restrict network access to vLLM HTTP and gRPC service endpoints to trusted networks and authenticated clients only. In host-based environments, restrict the serving port (typically TCP port 8000) using firewalld: ``` firewall-cmd --permanent --zone=trusted --add-source=<trusted_ip_or_subnet> firewall-cmd --permanent --zone=public --remove-port=8000/tcp firewall-cmd --reload ``` In Red Hat OpenShift AI deployments, apply a Kubernetes NetworkPolicy to the model-serving namespace to ensure only authorized ingress controllers or trusted application pods can reach the inference containers. Warning: Reloading firewall rules with `firewall-cmd --reload` applies changes immediately and may interrupt active network connections. Caveats: Network-level controls limit exposure to external untrusted callers, but they do not protect against crafted requests submitted by authorized users or internal services that consume the inference API.
- no fix planned: Will not fix
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
registry.redhat.io/rhoai/odh-ai-gateway-payload-processing-rhel9@sha256:1edfd89a924a83901845cd773012a3aeca70c09a2175a87e83351140efc80485_ppc64le as a component of Red Hat OpenShift AI 3.4 · known not affected

Vendor assessment for this product only.
Reason: vulnerable code not present

pkg:oci/odh-ai-gateway-payload-processing-rhel9@sha256%3A1edfd89a924a83901845cd773012a3aeca70c09a2175a87e83351140efc80485?arch=ppc64le&repository_url=registry.redhat.io/rhoai/odh-ai-gateway-payload-processing-rhel9&tag=1791860396
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
registry.redhat.io/rhoai/odh-eval-hub-rhel9@sha256:cb8f1ce5d2c2ae7a10d841837c7b18d3856f66869fc45d7bc8b4b193a4edbca8_arm64 as a component of Red Hat OpenShift AI 3.4 · known not affected

Vendor assessment for this product only.
Reason: vulnerable code not present

pkg:oci/odh-eval-hub-rhel9@sha256%3Acb8f1ce5d2c2ae7a10d841837c7b18d3856f66869fc45d7bc8b4b193a4edbca8?arch=arm64&repository_url=registry.redhat.io/rhoai/odh-eval-hub-rhel9&tag=1791852583
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
registry.redhat.io/rhoai/odh-llama-stack-core-rhel9@sha256:0aa47e4978f9ff14ca7e1341d4b26621cb8c271f0dff6fc8324c72afa5dbd17b_arm64 as a component of Red Hat OpenShift AI 3.4 · known not affected

Vendor assessment for this product only.
Reason: vulnerable code not present

pkg:oci/odh-llama-stack-core-rhel9@sha256%3A0aa47e4978f9ff14ca7e1341d4b26621cb8c271f0dff6fc8324c72afa5dbd17b?arch=arm64&repository_url=registry.redhat.io/rhoai/odh-llama-stack-core-rhel9&tag=1790958470
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
registry.redhat.io/rhoai/odh-maas-controller-rhel9@sha256:e305c12082dbed64798cc8ce855362f5703559f141308a7a0523f94d13886107_ppc64le as a component of Red Hat OpenShift AI 3.4 · known not affected

Vendor assessment for this product only.
Reason: vulnerable code not present

pkg:oci/odh-maas-controller-rhel9@sha256%3Ae305c12082dbed64798cc8ce855362f5703559f141308a7a0523f94d13886107?arch=ppc64le&repository_url=registry.redhat.io/rhoai/odh-maas-controller-rhel9&tag=1791790095
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
registry.redhat.io/rhoai/odh-mod-arch-autorag-rhel9@sha256:72b8f8d8a63c0664a729ec7afbfaa3e9ddb41d299253ed110b7358a91f55d144_ppc64le as a component of Red Hat OpenShift AI 3.4 · known not affected

Vendor assessment for this product only.
Reason: vulnerable code not present

pkg:oci/odh-mod-arch-autorag-rhel9@sha256%3A72b8f8d8a63c0664a729ec7afbfaa3e9ddb41d299253ed110b7358a91f55d144?arch=ppc64le&repository_url=registry.redhat.io/rhoai/odh-mod-arch-autorag-rhel9&tag=1791860994
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
registry.redhat.io/rhoai/odh-model-performance-data-rhel9@sha256:4ce3053cb04fdcdb6e2dfc5ba8f51cbd78d0d32c83f9e66ce376d273a592080f_ppc64le as a component of Red Hat OpenShift AI 3.4 · known not affected

Vendor assessment for this product only.
Reason: vulnerable code not present

pkg:oci/odh-model-performance-data-rhel9@sha256%3A4ce3053cb04fdcdb6e2dfc5ba8f51cbd78d0d32c83f9e66ce376d273a592080f?arch=ppc64le&repository_url=registry.redhat.io/rhoai/odh-model-performance-data-rhel9&tag=1790958081
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
registry.redhat.io/rhoai/odh-pipeline-runtime-minimal-cpu-py312-rhel9@sha256:a320d0547d3267d7ce0a3e004a44aed47d91b885ffe9840a65715e32a08ae0ec_arm64 as a component of Red Hat OpenShift AI 3.4 · known not affected

Vendor assessment for this product only.
Reason: vulnerable code not present

pkg:oci/odh-pipeline-runtime-minimal-cpu-py312-rhel9@sha256%3Aa320d0547d3267d7ce0a3e004a44aed47d91b885ffe9840a65715e32a08ae0ec?arch=arm64&repository_url=registry.redhat.io/rhoai/odh-pipeline-runtime-minimal-cpu-py312-rhel9&tag=1791829069
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
registry.redhat.io/rhoai/odh-th06-cuda130-torch210-py312-rhel9@sha256:7e0fb9bebdad1ebdef398243022fbfb29f6bd5673c00c4635745e0a92545ab60_arm64 as a component of Red Hat OpenShift AI 3.4 · known not affected

Vendor assessment for this product only.
Reason: vulnerable code not present

pkg:oci/odh-th06-cuda130-torch210-py312-rhel9@sha256%3A7e0fb9bebdad1ebdef398243022fbfb29f6bd5673c00c4635745e0a92545ab60?arch=arm64&repository_url=registry.redhat.io/rhoai/odh-th06-cuda130-torch210-py312-rhel9&tag=1791961113
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
registry.redhat.io/rhoai/odh-trustyai-vllm-orchestrator-gateway-rhel9@sha256:7b75fd02b3628b485fa58e89db548116c941941ec5da885767357e752bd03dd5_ppc64le as a component of Red Hat OpenShift AI 3.4 · known not affected

Vendor assessment for this product only.
Reason: vulnerable code not present

pkg:oci/odh-trustyai-vllm-orchestrator-gateway-rhel9@sha256%3A7b75fd02b3628b485fa58e89db548116c941941ec5da885767357e752bd03dd5?arch=ppc64le&repository_url=registry.redhat.io/rhoai/odh-trustyai-vllm-orchestrator-gateway-rhel9&tag=1791363252
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
rhoai/odh-th-torch-cuda-py312-rhel9 as a component of Red Hat OpenShift AI (RHOAI) · known not affected

Vendor assessment for this product only.
Reason: vulnerable code not present

pkg:oci/odh-th-torch-cuda-py312-rhel9@sha256:efd7b3607b699092c68359d20ef0d0043114191a097637667c0d2cf78fbea702?repository_url=registry.redhat.io/rhoai/odh-th-torch-cuda-py312-rhel9
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00

Known Affected Software Configurations

Vendor	Product	Version range
Vllm	Vllm	≥ 0.22.0 and < 0.24.0

All CVSS metrics

HIGH 8.2 v4.0 · CNA Primary
CVSS:4.0/AV:N/AC:H/AT:P/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N
MEDIUM 5.9 v3.1 · CNA Primary
CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:H
HIGH 8.2 v4.0 · NVD Secondary
CVSS:4.0/AV:N/AC:H/AT:P/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X
MEDIUM 5.9 v3.1 · NVD Secondary
CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:H
HIGH 7.5 v3.1 · NVD Primary
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H

References

Exploit-tagged references (not proof of a PoC)

https://github.com/vllm-project/vllm/security/advisories/GHSA-qff2-492f-9fm4 vendor-advisoryExploitMitigationVendor AdvisoryEVIDENCEvendor-referenceCVE record · VulnCheck · Updated 2026-09-26T13:23:21.865Z
NVD · CVE-2026-100652 · Updated 2026-10-06T19:22:01.530
PyPA · PYSEC-2026-4185
PyPA · PYSEC-2026-4185 · Updated 2026-10-07T10:00:03.378975746Z
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00

Advisories

https://www.vulncheck.com/advisories/vllm-0.22.0-through-0.23.0-denial-of-service-via-stop-token-ids third-party-advisoryThird Party AdvisoryADVISORYvendor-referenceCVE record · VulnCheck · Updated 2026-09-26T13:23:21.865Z
NVD · CVE-2026-100652 · Updated 2026-10-06T19:22:01.530
PyPA · PYSEC-2026-4185
PyPA · PYSEC-2026-4185 · Updated 2026-10-07T10:00:03.378975746Z
Red Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
https://osv.dev/vulnerability/PYSEC-2026-4185 ADVISORYPyPA · PYSEC-2026-4185
PyPA · PYSEC-2026-4185 · Updated 2026-10-07T10:00:03.378975746Z
https://security.access.redhat.com/data/csaf/v2/vex/2026/cve-2026-100652.json vendor-advisoryRed Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
https://euvd.enisa.europa.eu/vulnerability/EUVD-2026-87746 ADVISORYEUVD · EUVD-2026-87746

Technical & other

https://access.redhat.com/security/cve/CVE-2026-100652 vendor-referenceRed Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
https://bugzilla.redhat.com/show_bug.cgi?id=2541838 vendor-referenceRed Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
https://www.cve.org/CVERecord?id=CVE-2026-100652 vendor-referenceRed Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
https://nvd.nist.gov/vuln/detail/CVE-2026-100652 vendor-referenceRed Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00
https://access.redhat.com/errata/RHSA-2026:84734 vendor-remediationRed Hat · CVE-2026-100652 · Updated 2026-10-19T11:32:28+00:00

Sources: NVD · CVE.org · EPSS