---
id: CVE-2026-90777
title: >-
  ESPnet before 202609 deserializes pretrained model checkpoints using
  torch.load with weights_only=False, allowing arbitrary code execution from
  attacker-supplied files
summary: >-
  ESPnet before 202609 deserializes pretrained model checkpoints using
  torch.load with weights_only=False, allowing arbitrary code execution from
  attacker-supplied files. Attackers can craft malicious checkpoint files that
  execute code dur…
severity: high
cvss: 8.8
cvssVector: 'CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H'
cwe:
  - CWE-502
vendor: espnet
product: espnet
affected:
  - espnet < 202609
published: '2026-09-13'
updated: '2026-09-23'
sourceUpdated: '2026-09-23T17:17:44.357'
source: NVD
sourceUrl: 'https://nvd.nist.gov/vuln/detail/CVE-2026-90777'
references:
  - url: 'https://github.com/espnet/espnet'
    label: disclosure@vulncheck.com
  - url: >-
      https://github.com/espnet/espnet/blob/v.202511/espnet2/torch_utils/load_pretrained_model.py#L99
    label: disclosure@vulncheck.com
  - url: >-
      https://github.com/espnet/espnet/commit/91ca045fc179f29bc7b7fe05cd852aea65310003
    label: disclosure@vulncheck.com
  - url: 'https://github.com/espnet/espnet/releases/tag/v.202609'
    label: disclosure@vulncheck.com
  - url: 'https://github.com/espnet/espnet/security/advisories/GHSA-64f6-3gqc-r926'
    label: disclosure@vulncheck.com
  - url: >-
      https://www.vulncheck.com/advisories/espnet-before-202609-remote-code-execution-via-unsafe-deserialization
    label: disclosure@vulncheck.com
  - url: 'https://github.com/espnet/espnet/security/advisories/GHSA-64f6-3gqc-r926'
    label: 134c704f-9b21-4f2e-91b3-4a467353bcc0
tags:
  - nvd
  - cve.org
  - exploit-available
epss: 0.00507
epssPercentile: 0.42341
exploitAvailable: true
ssvc:
  exploitation: poc
  automatable: 'no'
  technicalImpact: total
  timestamp: '2026-09-14T17:27:19.541240Z'
ingestedAt: '2026-09-14T15:23:07.470Z'
---

## Overview

ESPnet before 202609 deserializes pretrained model checkpoints using torch.load with weights_only=False, allowing arbitrary code execution from attacker-supplied files. Attackers can craft malicious checkpoint files that execute code during deserialization when loaded through the initialization or fine-tuning path.

## Remediation

Refer to the linked advisories for vendor-supplied fixes and affected version ranges.