(0Day) Ultralytics torch_load Deserialization of Untrusted Data Remote Code Execution (CVE‑2026‑118613)
What It Is — A zero‑day vulnerability (CVE‑2026‑118613) in the torch_load function of Ultralytics’ AI/ML library allows a remote attacker to execute arbitrary code on the host running the library. The flaw stems from deserialization of attacker‑controlled data without proper validation. Exploitation requires user interaction (visiting a malicious page or opening a crafted file), after which the attacker’s payload runs with the privileges of the current user. The advisory, published by the Zero Day Initiative on 28 Oct 2026, assigns a CVSS 7.8 (High) score (AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H). No public patch is yet available; mitigation is limited to restricting interaction with the vulnerable component.
Affected Products — Ultralytics AI/ML library (the torch_load function) used in Python environments.
Why It Matters for Trust & Control Assurance — The issue tests the control objective of secure development and input validation—a single control that maps to many frameworks (e.g., NIST CSF 2.0, ISO 27001). Weaknesses in deserialization expose organizations to RCE, undermining audit‑ready evidence of safe coding practices.
Relevant Verisq capability: Control‑Mapping (continuous evidence collection and mapping of code‑level controls).
Recommended Actions
- Restrict execution of untrusted code and limit exposure of the Ultralytics library to unverified inputs.
- Deploy any vendor‑released patches or updates as soon as they become available.
- Implement runtime monitoring to detect anomalous process creation or file writes originating from
torch_load. - Conduct a code‑review focused on deserialization paths and add unit tests that enforce strict input validation.
- Document the remediation steps as audit evidence for secure‑development controls.