AI Stack Alert Medium TD-2026-0071 For AI Eng, AppSec, SOC / IR AI-assisted
Meta fairseq FConvModel YAML parsing flaw allows remote code execution via malicious file
ZDI published advisory ZDI-26-941 describing a code injection flaw in Meta fairseq's configuration file parsing. The flaw stems from unvalidated user-supplied strings used to execute Python code, and exploitation requires a target to open a malicious file. The vendor stated the fairseq repository is no longer maintained and was archived.
Why it matters
Organizations running fairseq can have arbitrary code executed in the context of the current process if a user opens a crafted file. Because the repository is archived, the vendor is unlikely to issue a fix.
Do this
- Identify systems and ML pipelines that run fairseq, including FConvModel-related workloads.
- Restrict users and automated jobs from opening untrusted model configuration or YAML files with fairseq.
- Run fairseq workloads in isolated, least-privilege environments to limit impact of code execution.
- Plan migration away from the archived fairseq repository to a maintained alternative.
- Monitor for unexpected Python child processes or outbound connections from fairseq hosts.
Signals
- NoCISA KEV
- 24/100Priority
- 60%Confidence
Evidence
“The issue results from the lack of proper validation of a user-supplied string before using it to execute Python code.”
“the vendor communicated that the fairseq repository is no longer maintained and was archived”
“User interaction is required to exploit this vulnerability in that the target must open a malicious file.”
Sources
Affected / related
Meta fairseq, YAML configuration parsing code injection
Published 2026-11-11 06:00 UTC · updated 2026-11-12 17:35 UTC · AI-assisted summary, checked against quoted sources · How we score