How severe is CVE-2026-80047?

CVE-2026-80047 has a CVSS 3.x base score of 7.8, rated high severity. It is exploitable over local access with low attack complexity, requires low privileges and no user interaction. Impact on confidentiality is high, integrity high, and availability high.

Is CVE-2026-80047 being actively exploited?

It is not currently listed in CISA's KEV catalog. Its EPSS exploit-prediction score is 0% (1st percentile), an estimate of the probability of exploitation in the next 30 days.

How do I fix CVE-2026-80047?

Review the linked vendor and NVD advisories for patched versions and mitigations, then upgrade or apply the recommended workaround. Given its high severity, prioritise patching exposed systems.

When was CVE-2026-80047 published?

CVE-2026-80047 was published on 2026-09-01 and last updated on 2026-09-23.

CVE-2026-80047

CVE-2026-80047 is a high-severity vulnerability with a CVSS 3.x base score of 7.8. It is not currently listed as actively exploited by CISA, and its EPSS exploit-prediction score is low. The underlying weakness is classified as CWE-273.

Key facts

Severity: High (CVSS 3.x base score 7.8)
EPSS exploit prediction: 0% (1st percentile)
Actively exploited: Not listed in CISA KEV
Weakness: CWE-273
Published: 2026-09-01
Last modified: 2026-09-23

Description

A vulnerability in Hugging Face Transformers (versions 4.57.0 to 5.16.1) allows remote Python files to be written to local disk without user consent when using GenerativePreTrainedModel.load_custom_generate(). The function fetches and caches a remote module file before performing the required trust_remote_code consent check, inverting the security model enforced by other code-loading paths (such as AutoConfig, AutoModel, and AutoTokenizer). As a result, attacker‑controlled Python code from custom_generate/generate.py is copied into the user’s ~/.cache/huggingface/modules directory even if the user declines the trust prompt. Although execution is correctly gated, the file write is not reversible and can persist across sessions. This can lead to persistent, unauthorized files on disk and stale cache collisions where cached attacker code may later be executed during trusted model loads. The issue stems from an unconditional file write in dynamic_module_utils.py prior to any trust verification.

Frequently asked questions

What is CVE-2026-80047?: A vulnerability in Hugging Face Transformers (versions 4.57.0 to 5.16.1) allows remote Python files to be written to local disk without user consent when using GenerativePreTrainedModel.load_custom_generate(). The function fetches and caches a remote module file before performing the required trust_remote_code consent check, inverting the security model enforced by other code-loading paths (such as AutoConfig, AutoModel, and AutoTokenizer). As a result, attacker‑controlled Python code from custom_generate/generate.py is copied into the user’s ~/.cache/huggingface/modules directory even if the user declines the trust prompt. Although execution is correctly gated, the file write is not reversible and can persist across sessions. This can lead to persistent, unauthorized files on disk and stale cache collisions where cached attacker code may later be executed during trusted model loads. The issue stems from an unconditional file write in dynamic_module_utils.py prior to any trust verification.
How severe is CVE-2026-80047?: CVE-2026-80047 has a CVSS 3.x base score of 7.8, rated high severity. It is exploitable over local access with low attack complexity, requires low privileges and no user interaction. Impact on confidentiality is high, integrity high, and availability high.
Is CVE-2026-80047 being actively exploited?: It is not currently listed in CISA's KEV catalog. Its EPSS exploit-prediction score is 0% (1st percentile), an estimate of the probability of exploitation in the next 30 days.
How do I fix CVE-2026-80047?: Review the linked vendor and NVD advisories for patched versions and mitigations, then upgrade or apply the recommended workaround. Given its high severity, prioritise patching exposed systems.
When was CVE-2026-80047 published?: CVE-2026-80047 was published on 2026-09-01 and last updated on 2026-09-23.

References

Other CWE-273 vulnerabilities

CVE-2015-0278 — Critical (CVSS 10.0): libuv before 0.10.34 does not properly drop group privileges, which allows context-dependent attackers to gain…
CVE-2023-34844 — Critical (CVSS 9.8): Play With Docker < 0.0.2 has an insecure CAP_SYS_ADMIN privileged mode causing the docker container to escape.
CVE-2021-36372 — Critical (CVSS 9.8): In Apache Ozone versions prior to 1.2.0, Initially generated block tokens are persisted to the metadata database and…
CVE-2020-24361 — Critical (CVSS 9.8): SNMPTT before 1.4.2 allows attackers to execute shell code via EXEC, PREXEC, or unknown_trap_exec.
CVE-2011-3350 — Critical (CVSS 9.8): masqmail 0.2.21 through 0.2.30 improperly calls seteuid() in src/log.c and src/masqmail.c that results in improper…
CVE-2011-2921 — Critical (CVSS 9.8): ktsuss versions 1.4 and prior has the uid set to root and does not drop privileges prior to executing user specified…

Browse all CWE-273 vulnerabilities →