AI Stack Alert High CVE-2026-126974 For AppSec, AI Eng, SOC / IR AI-assisted
llama.cpp GGUF file parsing integer overflow allows remote code execution (CVE-2026-126974)
A remote attacker can execute arbitrary code on affected llama.cpp installations. The flaw is in GGUF file parsing, where missing validation of user-supplied data causes an integer overflow before a memory write. User interaction is required: the target must visit a malicious page or open a malicious file.
Why it matters
Any user or service that loads untrusted GGUF model files with an affected llama.cpp build could have code run in the context of that process. Developer workstations and inference servers that load third-party models are most exposed.
Do this
- Identify llama.cpp deployments and builds that parse GGUF files, including developer workstations and inference servers.
- Update llama.cpp to a build that includes the fix referenced in upstream pull request 31336.
- Restrict loading of GGUF files to trusted sources until patched, and block untrusted model downloads.
- Monitor for unexpected child processes or memory-corruption crashes from llama.cpp or tools that embed it.
Exploit Clock
No signals yet
- No exploitation signals yet
Signals
- NoCISA KEV
- 34/100Priority
- 70%Confidence
Evidence
“The specific flaw exists within the parsing of GGUF files. The issue results from the lack of proper validation of user-supplied data, which can result in an integer overflow before writing to memory.”
“User interaction is required to exploit this vulnerability in that the target must visit a malicious page or open a malicious file.”
“llama.cpp has issued an update to correct this vulnerability.”
Sources
Affected / related
llama.cpp, GGUF file parser
Published 2026-11-13 06:00 UTC · updated 2026-11-13 21:35 UTC · AI-assisted summary, checked against quoted sources · How we score