CVE-2026-5817
Docker Model Runner container-to-host code execution via unsandboxed trust_remote_code in Python inference backends
No known exploitation. EPSS puts it in the 7th percentile. Only a workaround so far.
What to do
No fixed build is published yet. The vendor describes a workaround.
What it is
From the CVE record
The vllm-metal inference backend in Docker Model Runner on macOS unconditionally sets trust_remote_code=True when loading model tokenizers, and runs without sandboxing. This causes transformers.AutoTokenizer.from_pretrained() to import and execute arbitrary Python files included in any model pulled from an OCI registry, resulting in arbitrary code execution on the Docker host as the Docker Desktop user when inference is triggered. Any container on the Docker network can trigger this by calling the model-runner.docker.internal API to pull a malicious model and request inference.
In plain language
No plain-language summary for this CVE yet.
Exploitation
Where each signal puts this CVE on the scale from published to confirmed exploited.
- CISA KEV
Not in the catalog. CISA has not confirmed exploitation.
- Public exploits
No public exploit or proof of concept found in the sources we track.
- EPSS
0.2% chance of exploitation activity in the next 30 days, which ranks it in the 7th percentile of scored CVEs.
Exploit Prediction Scoring System, FIRST.org. A probability, not a confirmation.
Lifecycle
8 events over 62 days, from the signal feeds we watch.
- Workaround availablerecord updated
- OpenVAS check added
- Workaround availablerecord updated
- Publishedweakness classified, att&ck mapped
Affected products
And 1 more affected product. See all after sign-in
Technical detail
CVSS 3.1 vector
Open in the CVSS calculatorCVSS:3.1/AV:L/AC:L/PR:L/UI:R/S:C/C:H/I:H/A:H
Scored 8.2 by NVD.
How it is reached
- Attack Vector LocalRequires local access to the vulnerable system (e.g. local login, malicious file)
- Attack Complexity LowNo special conditions — the attack can be reliably reproduced
- Privileges Required LowRequires basic user-level privileges
- User Interaction RequiredA user must click a link, open a file, or perform some action
Scope
- Scope ChangedThe exploit can affect other components (e.g. sandbox escape, host compromise from VM)
Impact if exploited
- Confidentiality HighTotal information disclosure — all data in the component is compromised
- Integrity HighTotal loss of integrity — attacker can modify any data in the component
- Availability HighTotal denial of service — the component is completely unavailable
Weaknesses
ATT&CK techniques
Mapped from the weaknesses above (CWE to ATT&CK), not observed in attacks.
- Initial AccessT1195Supply Chain Compromisehigh confidence
Sources
References in the record
Watch the software you run.
My Stack ranks new CVEs for your products by real-world exploitation, so the next serious one reaches you without reading every advisory.
We'll flag the next CVE, public exploit or patch for Docker Desktop, not every advisory.
A free account adds
- The full version matrix and every affected product
- Exploit links, proofs of concept and Metasploit modules
- Email alerts for the products you watch
- The same data over REST API, MCP and CLI