Llama.Cpp
CVE-2026-21869
HIGH
Attack ChainAIDerived
Hypothetical attack flow derived from CVE metadata – not a confirmed exploit chain
-
AccessSend completion request with negative n_discard parameter
-
DeliveryParse JSON input without validation
-
ExploitTrigger reversed range in llama_memory_seq_rm
-
ExecutionWrite out-of-bounds memory in token loop
-
ImpactCrash process or execute arbitrary code
Vulnerability Assessment
| Weakness |
CWE
OWASP Top 10:2025
A06:2021 - Vulnerable and Outdated Components
ASVS 5.0
V15 - Secure Coding and Architecture Bounds checking, write limits
MITRE ATT&CK
2 techniques across 2 tactics: Execution, Privilege Escalation The technique list with ATT&CK links and the kill-chain view is part of the Basic plan and up. |
| Exploitation Conditions | llama.cpp server with completion endpoints enabled. Remote attacker must send crafted JSON request with negative n_discard value when context buffer is full. No authentication required. Affects commits 55d4206c8 and prior without patch. |
| Impact | CVSS 8.8 (HIGH). Attack vector: Network. Complexity: Low. Privileges required: None. User interaction: Required. Exploitable via remote/network. Public exploit/POC available. |
| Exploit ScenarioDerived | A remote attacker could exploit this vulnerability to compromise the affected system. |
Get an e-mail when this CVE is exploited, patched or re-scored. Free account.
Alert meAnalysisAIv1 – Mar 12, 2026
Llama.cpp server endpoints fail to validate the n_discard parameter from JSON input, allowing negative values that trigger out-of-bounds memory writes when the context buffer fills. This memory corruption vulnerability affects LLM inference operations and can be exploited remotely without authentication to crash the service or achieve code execution; public exploit code exists and no patch is currently available.
Earlier versions of this analysis (1)
llama.cpp is an inference of several LLM models in C/C++. In commits 55d4206c8 and prior, the n_discard parameter is parsed directly from JSON input in the llama.cpp server's completion endpoints without validation to ensure it's non-negative. [CVSS 8.8 HIGH]
RemediationAI
Monitor vendor advisories for a patch. Enable ASLR, DEP/NX, and stack canaries where possible. Restrict network access to the affected service where possible.
Affects your stack
Find out whether the stacks and servers you run are on an affected version.
Checking this CVE against your own stacks and servers is part of the Basic plan and up.
Technical ContextAI · show details
Classified as CWE-787 (Out-of-bounds Write). Affects Llama.Cpp. llama.cpp is an inference of several LLM models in C/C++. In commits 55d4206c8 and prior, the n_discard parameter is parsed directly from JSON input in the llama.cpp server's completion endpoints without validation to ensure it's non-negative. When a negative value is supplied and the context fills up, llama_memory_seq_rm/add receives a reversed range and negative offset, causing out-of-bounds memory writes in the token evaluation loop. This deterministic memory corruption can crash the process
Know when Llama.Cpp gets its next vulnerability
- Watch products and get an email when something new is published
- Check this CVE against your own stack and servers
- Exploit and PoC references, threat intelligence from 22 sources
No credit card · 7-day full trial
Severity by source
AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H
Primary rating from GitHub Advisory.
CVSS VectorGitHub Advisory
Lifecycle Timeline
2DescriptionGitHub Advisory
llama.cpp is an inference of several LLM models in C/C++. In commits 55d4206c8 and prior, the n_discard parameter is parsed directly from JSON input in the llama.cpp server's completion endpoints without validation to ensure it's non-negative. When a negative value is supplied and the context fills up, llama_memory_seq_rm/add receives a reversed range and negative offset, causing out-of-bounds memory writes in the token evaluation loop. This deterministic memory corruption can crash the process or enable remote code execution (RCE). There is no fix at the time of publication.
Executive ViewAI
Organizations using llama.cpp face significant risk from CVE-2026-21869, a out-of-bounds write vulnerability rated CVSS 8.8. Exploitation could allow attackers to corrupt memory, crash the application, or potentially execute arbitrary code. Proof-of-concept code is publicly available, increasing the risk of exploitation. Organizations should apply vendor updates when available and consider compensating controls in the interim.
Apply vendor patch. If immediate patching is not feasible: (1) Enable ASLR and DEP where available; (2) Implement network segmentation; (3) Monitor for exploitation indicators.
Threat Intelligence Overview
NVD
CVE.org
EPSS
GitHub
How this decision was reached
The engine reads each signal below and applies a fixed list of rules from top to bottom. The first rule that matches sets the recommended action – like a statement, every line is shown.
What every action level means
SSVC AssessmentENISA
Priority Score
Attack Technique
Memory corruption occurs when a program writes data beyond the boundaries of allocated memory regions or accesses memory in unintended ways, violating the integrity of the process's address space.
Remote Code Execution represents the critical moment when an attacker successfully runs arbitrary code on a target system without physical access.
Denial of Service attacks render applications or systems unavailable by overwhelming resources or triggering failure conditions.
Affected Products (CPE)GitHub Advisory
CVE-2025-52566 is a signed vs. unsigned integer overflow vulnerability in llama.cpp's tokenizer (llama_vocab::tokenize f
A buffer overflow vulnerability in llama.cpp (CVSS 8.8). High severity vulnerability requiring prompt remediation. Vendo
Exploit-DB