BlackTreeIndependent security intelligence
← Back to the CVE catalogue
Full vulnerability report · 2026
CVE-2026-44513High confidence

Diffusers: `trust_remote_code` bypass via `custom_pipeline` and local custom components

huggingface · diffusers

8.8HighCVSS 3.1
Recommended action
Within 7 days

High technical severity with public exploit material referenced by a structured source; prioritise exposed affected systems while verifying vendor guidance.

Patch available
Direct vendor intelligence

Authoritative vendor CSAF and VEX advisories

Structured product status and remediation from the issuing vendor. Product-state explanations are always visible; large lists can be searched or downloaded.

1 current
CVE-2026-44513 · CSAF 2.0 · revision 3 · finalRed Hat Product SecurityDiffusers: Diffusers: Arbitrary remote code execution via `trust_remote_code` bypass
9 known affected

The vendor explicitly identifies these products as affected by this CVE.

  • rhaii/vllm-cpu-rhel9 as a component of Red Hat AI Inference Server
  • rhaii/vllm-cuda-rhel9 as a component of Red Hat AI Inference Server
  • rhaiis/vllm-cpu-rhel9 as a component of Red Hat AI Inference Server
  • rhaiis/vllm-tpu-rhel9 as a component of Red Hat AI Inference Server
  • rhelai3/bootc-aws-cuda-rhel9 as a component of Red Hat Enterprise Linux AI (RHEL AI) 3
  • rhelai3/bootc-azure-cuda-rhel9 as a component of Red Hat Enterprise Linux AI (RHEL AI) 3
  • rhelai3/bootc-cuda-rhel9 as a component of Red Hat Enterprise Linux AI (RHEL AI) 3
  • rhelai3/bootc-gcp-cuda-rhel9 as a component of Red Hat Enterprise Linux AI (RHEL AI) 3
  • rhoai/odh-openvino-model-server-rhel9 as a component of Red Hat OpenShift AI (RHOAI)
Summary
A flaw was found in Diffusers, a library for pretrained diffusion models. A remote attacker could exploit a bypass in the `trust_remote_code` mechanism within the `DiffusionPipeline.from_pretrained` function. This vulnerability allows for arbitrary remote code execution, even when the user explicitly sets `trust_remote_code=False` or omits it. The issue stems from the security check being incorrectly placed, allowing malicious code to be loaded and executed by bypassing the intended security gate.
Remediation
For Red Hat OpenShift AI 3.4.4 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/
Optional official sources

National CERT insights
?CERT means Computer Emergency Response Team; CSIRT is the closely related term Computer Security Incident Response Team.

Select the national-authority views to include. The exact source language is shown on each matched advisory. Your choice is remembered on this device and encoded in the shareable URL.

Official European source

ENISA European Vulnerability Database

Official EUVD identifiers, advisory evidence and known-exploited context. Missing fields are not treated as evidence of low risk.

1 current
ENISA EUVD identifier

EUVD-2026-30334

No EUVD known-exploited evidence

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.

EUVD state
Present in the current official mapping
Known exploitation
Not present in the current ENISA EUVD known-exploited dataset. This is not proof of no exploitation.
ENISA score
8.8 · CVSS 3.1
Advisory evidence
No linked advisory details stored yet
Recommended actionWithin 7 days

High technical severity with public exploit material referenced by a structured source; prioritise exposed affected systems while verifying vendor guidance.

Patch available
01

What, why and how

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.

What

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.

Why

Untrusted data can cross into a code-evaluation path and be interpreted as executable instructions.

How

An attacker operating through a network path may attempt exploitation without authentication after a user interaction. If successful, the issue may execute code or commands in the affected security context.

What

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.

Why

Untrusted data can cross into a code-evaluation path and be interpreted as executable instructions.

How

An attacker operating through a network path may attempt exploitation without authentication after a user interaction. If successful, the issue may execute code or commands in the affected security context.

02

Exploit reality and attack path

CVSS severity, EPSS forecast probability, public exploit material and CISA-confirmed exploitation are separate signals.

Observed exploitation
?Confirmed exploitation and public exploit material are separate signals. Attacks can occur without public proof-of-concept or exploit code.
No confirmed evidence

No CISA KEV match was present at the last successful refresh. This means no confirmation from that source, not proof of no exploitation.

Public PoC / exploit material
?Confirmed exploitation and public exploit material are separate signals. Attacks can occur without public proof-of-concept or exploit code.
Reference recorded

A structured CVE source labels at least one public reference as exploit material. BlackTree has not independently validated that it is safe, reliable or weaponised.

Likely attack path
a network path → Improper Control of Generation of Code ('Code Injection') → execute code or commands in the affected security context
Attack surface
Network
Privileges required
None: unauthenticated exploitation is possible
User interaction
Required interaction required
Attack complexity
Low: no specialised conditions are recorded
Security boundary
Unchanged: impact remains within the vulnerable component's security authority
Weakness
?CWE means Common Weakness Enumeration: a standard category for the underlying weakness.
CWE-94 ↗

CWE-94: Improper Control of Generation of Code ('Code Injection'). The product constructs all or part of a code segment using externally-influenced input from an upstream component, but it does not neutralize or incorrectly neutralizes special elements that could modify the syntax or behavior of the intended code segment.

CVSS vector
?CVSS means Common Vulnerability Scoring System. The vector records the metric values used to calculate technical severity.
CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H

Common Vulnerability Scoring System 3.1: the compact vector below is decoded into plain language.

AVNetworkAttack vector: The vulnerable component can be reached over a network.ACLowAttack complexity: No specialised conditions are required beyond attacker-controlled input.PRNonePrivileges required: The attacker does not need an account or existing privileges.UIRequiredUser interaction: Another user must perform an action for exploitation to succeed.SUnchangedScope: The security impact remains within the vulnerable component's authority.CHighConfidentiality impact: A successful attack can cause a major loss.IHighIntegrity impact: A successful attack can cause a major loss.AHighAvailability impact: A successful attack can cause a major loss.
Post-exploitation / living off the land
After compromise, an attacker may use built-in shells, scripting engines, scheduled tasks and native network utilities for discovery, persistence or movement. This is a plausible LoTL path, not evidence that it has occurred for every attack.
NetworkUnauthenticatedRemote code executionCWE-94Public exploit reference
A

Official authority intelligence

Only matched European and national findings are included. Language selectors and unavailable sources are omitted.

ENISA EUVD · EUVD-2026-30334Official EUVD mapping

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.

Official EUVD record
03

Patch and workaround

Operational remediation based on structured source evidence.

Status
?Patch availability is based on structured fixed-version fields and authoritative update references. If no fix is verified, check the vendor advisory before making a change.
Patch available
Affected
rhaii/vllm-cpu-rhel9 as a component of Red Hat AI Inference Server; rhaii/vllm-cuda-rhel9 as a component of Red Hat AI Inference Server; rhaiis/vllm-cpu-rhel9 as a component of Red Hat AI Inference Server; rhaiis/vllm-tpu-rhel9 as a component of Red Hat AI Inference Server; rhelai3/bootc-aws-cuda-rhel9 as a component of Red Hat Enterprise Linux AI (RHEL AI) 3; rhelai3/bootc-azure-cuda-rhel9 as a component of Red Hat Enterprise Linux AI (RHEL AI) 3; rhelai3/bootc-cuda-rhel9 as a component of Red Hat Enterprise Linux AI (RHEL AI) 3; rhelai3/bootc-gcp-cuda-rhel9 as a component of Red Hat Enterprise Linux AI (RHEL AI) 3; rhoai/odh-openvino-model-server-rhel9 as a component of Red Hat OpenShift AI (RHOAI)
Fixed
registry.redhat.io/rhoai/odh-th06-cuda130-torch210-py312-rhel9@sha256:5e8cac66dc108c958173412601f1c37e777f53209ff9166add58a011823c87f6_arm64 as a component of Red Hat OpenShift AI 3.4; registry.redhat.io/rhoai/odh-th06-cuda130-torch210-py312-rhel9@sha256:ea69ab01a3aaeba9b4db8abbace9014a3044e0f0f13fb7aaaef93acf72f50248_amd64 as a component of Red Hat OpenShift AI 3.4; registry.redhat.io/rhoai/odh-th06-rocm64-torch291-py312-rhel9@sha256:b2a253f3dc063a8e1cee441640cce13fcf04918ce453be8f5d488301b1458a91_amd64 as a component of Red Hat OpenShift AI 3.4; registry.redhat.io/rhoai/odh-training-cuda128-torch29-py312-rhel9@sha256:d5cc3e7ad03f83104f3d1d4038fb2a4fa11dfa7501ce6be28ac568459a56f089_amd64 as a component of Red Hat OpenShift AI 3.4; registry.redhat.io/rhoai/odh-training-rocm64-torch29-py312-rhel9@sha256:bd98eb46f3f5e676457617ca0bfa7117610abf9d69b6082cb37b9a880af26eb2_amd64 as a component of Red Hat OpenShift AI 3.4
Action
For Red Hat OpenShift AI 3.4.4 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/
Workaround
A mitigation or workaround reference is available from the source linked below; validate it against the affected product and version.
04

Evidence and provenance

Published 14 May 2026 · Last source change 28 Aug 2026, 12:04 UTC · CWE-94 · Improper Control of Generation of Code ('Code Injection')

CVE recordCVE.org · 5.2
CVSS sourceCNA
EPSS source
?The date BlackTree first stored a score for this CVE from the daily FIRST EPSS feed.
FIRST · tracked since 2026-08-14
European sourceENISA EUVD · EUVD-2026-30334
Product sourceVendor CSAF · Red Hat Product Security
Remediation sourceVendor CSAF · Red Hat Product Security
CWE sourceCNA
NVD statusNVD modified after enrichment

Core structured fields are present and their contributing authorities are shown above.

Material change intelligence

What changed after publication

View recent updates →
  1. Vendor guidanceAuthoritative vendor guidance changed from authoritative remediation link · mitigation, vendor advisory to authoritative remediation link · mitigation.
    Before
    authoritative remediation link · mitigation, vendor advisory
    After
    authoritative remediation link · mitigation
    security-advisories@github.com ↗
  2. Vendor guidanceAuthoritative vendor guidance changed from authoritative remediation link · mitigation to authoritative remediation link · mitigation, vendor advisory.
    Before
    authoritative remediation link · mitigation
    After
    authoritative remediation link · mitigation, vendor advisory
    security-advisories@github.com ↗
Material fields only · duplicate refreshes suppressed · history retained for the configured operational retention period
CVE published
Mitigation reference recorded
ENISA EUVD · EUVD-2026-30334↗Vendor CSAF · Red Hat Product Security↗Public exploit reference↗Official CVE record (CNA and ADP data)↗NVD record↗
Technical terms and abbreviations used in this report
CVE
Common Vulnerabilities and Exposures: the public identifier for one disclosed vulnerability.
CVSS
Common Vulnerability Scoring System: a technical severity framework; it is not patching priority by itself.
EPSS
Exploit Prediction Scoring System: FIRST's estimate of the probability that exploitation activity will be observed in the next 30 days; it is a forecast, not confirmation.
CWE
Common Weakness Enumeration: the standard category describing the underlying software or hardware weakness.
CNA
CVE Numbering Authority: an organisation authorised to assign and publish CVE records.
CISA ADP
Cybersecurity and Infrastructure Security Agency Authorized Data Publisher: structured enrichment added to a CVE record.
NVD
National Vulnerability Database: NIST's enrichment service for CVE records.
CERT / CSIRT
A computer security incident response team that publishes warnings or coordinates incident response.
PoC
Proof of concept: public material that demonstrates or helps reproduce exploitation.
CSAF
Common Security Advisory Framework: a machine-readable format for security advisories.
LoTL
Living off the land: abuse of legitimate tools or system functions during an attack.
Free version - for non-commercial use only.CVE-2026-44513 · cve.blacktree.nl