CVE-2025-9959

Affects: prompt injection

CVSS: HIGH · 7.6v3.1
CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:L/I:H/A:L
Published: 2025-09-03
Weakness: CWE-94
Source: nvd.nist.gov/vuln/detail/CVE-2025-9959

Description

Incomplete validation of dunder attributes allows an attacker to escape from the Local Python execution environment sandbox, enforced by smolagents. The attack requires a Prompt Injection in order to trick the agent to create malicious code.

References

https://github.com/huggingface/smolagents/pull/1551
https://research.jfrog.com/vulnerabilities/smolagents-local-python-sandbox-escape-jfsa-2025-001434277/

How GTK Cyber trains on this

AI security training at GTK Cyber covers the LLM and ML-pipeline vulnerability classes that vulnerabilities like CVE-2025-9959 fall into. Our hands-on courses are taught by Charles Givre and other practitioners who break and defend production AI systems.

AI Red-Teaming course →·Browse MITRE ATLAS techniques

Related AI/LLM CVEs

CVE-2026-10564(HIGH)
CVE-2026-55607(HIGH)
CVE-2026-48719(HIGH)
CVE-2025-66389(HIGH)
CVE-2024-58351(CRITICAL)
CVE-2026-12045(CRITICAL)
CVE-2026-22551(MEDIUM)
CVE-2026-44688(HIGH)

AI security training, taught by people who do the work.

Hands-on courses on adversarial AI, prompt injection, and ML pipeline security.

Explore AI Red-Teaming