Diffusers is the a library for pretrained diffusion models
Why It Matters
Analysis dossierWhy It Matters
Diffusers is the a library for pretrained diffusion models. Prior to 0.38.0, a trust_remote_code bypass in DiffusionPipeline.from_pretrained allows arbitrary remote code execution despite the user passing trust_remote_code=False (or omitting it, which is the default). The vulnerability has three variants, all sharing the same root cause — the trust_remote_code gate was implemented inside DiffusionPipeline.download() rather than at the actual dynamic-module load site, so any code path that bypassed or short-circuited download() also bypassed the security check. DiffusionPipeline.from_pretrained('repoA', custom_pipeline='attacker/repoB', trust_remote_code=False) — the gate evaluated against repoA's file list rather than repoB's, so repoB's pipeline.py was loaded and executed. DiffusionPipeline.from_pretrained('/local/snapshot', custom_pipeline='attacker/repoB', trust_remote_code=False) — the local-path branch never invoked download(), so the gate was never reached and remote code from repoB executed. DiffusionPipeline.from_pretrained('/local/snapshot', trust_remote_code=False) where the snapshot contains custom component files (e.g. unet/my_unet_model.py) referenced from model_index.json — same root cause; the local path skipped download() and custom component code executed. This vulnerability is fixed in 0.38.0.
High
Unknown
Unknown
Risk basis: SOC ecosystem technology relevance; HIGH severity; Source severity; no confirmed exploitation signal
Affected Products & Remediation
1 source-backed product- huggingface diffusers · < 0.38.0 · NVD CPE · Source
Vendor or catalog guidance (source)
Not provided by source
Known exploitation (source facts)
Only what a source states. Unknown means no source has said either way; it is not "not exploited".
- CISA KEV
- Unknown
- Exploited (source statement)
- Unknown
- Publicly disclosed (source)
- Unknown
Suggested hunting (rule-based inference)
SecOps Signal recommendation (rule-based inference). These steps are hunting hypotheses from the vulnerability class; they do not state or imply that exploitation occurred here or anywhere.
Matched user-interaction-client (CVSS AV:N, CVSS UI:R)
Exposure
- Identify endpoints reporting this CVE, especially those used for email and web browsing.Telemetry: DeviceTvmSoftwareVulnerabilities · A network vector that needs user interaction points at user-facing client software. · Rule user-interaction-client
Exploitation evidence
- Review unusual child processes started by the affected client application on affected endpoints.Telemetry: DeviceProcessEvents · Hunting hypothesis from the CVSS vector; it does not show that exploitation occurred. · Rule user-interaction-client
Response
- Confirm the source-stated update on affected endpoints.Telemetry: DeviceTvmSoftwareVulnerabilities · A follow-up inventory check verifies remediation. · Rule user-interaction-client
CISA KEV Record
CISA KEV status was not determined for this record.
Change History & Source Data
— No recorded material change
- CNA / source
- security-advisories@github.com
- Record status · verification
- UPDATED · NVD API
- CPE
- cpe:2.3:a:huggingface:diffusers:*:*:*:*:*:python:*:*
Evidence & Provenance
Linked by the query's stored exact CVE association. Review required tables and validation status before use.
DeviceTvmSoftwareVulnerabilities| where CveId == "CVE-2026-44513"| project DeviceId, DeviceName, OSPlatform, OSVersion, SoftwareVendor, SoftwareName, SoftwareVersion, CveId, VulnerabilitySeverityLevel, RecommendedSecurityUpdate, RecommendedSecurityUpdateId| sort by VulnerabilitySeverityLevel descLooking up related community queries…