CVE-2024-14021: CWE-502 Deserialization of Untrusted Data in run-llama llama_index
LlamaIndex (run-llama/llama_index) versions up to and including 0.11.6 contain an unsafe deserialization vulnerability in BGEM3Index.load_from_disk() in llama_index/indices/managed/bge_m3/base.py. The function uses pickle.load() to deserialize multi_embed_store.pkl from a user-supplied persist_dir without validation. An attacker who can provide a crafted persist directory containing a malicious pickle file can trigger arbitrary code execution when the victim loads the index from disk.
AI Analysis
Technical Summary
CVE-2024-14021 is a deserialization of untrusted data vulnerability in the run-llama llama_index project, specifically in the BGEM3Index.load_from_disk() method located in llama_index/indices/managed/bge_m3/base.py. The method uses Python's pickle.load() to deserialize the multi_embed_store.pkl file from a user-controlled persist_dir without validating the input. This allows an attacker who can control the persist_dir contents to craft a malicious pickle file that triggers arbitrary code execution upon loading the index. The vulnerability affects all versions up to and including 0.11.6.
Potential Impact
An attacker who can supply a malicious persist directory to the vulnerable function can achieve arbitrary code execution on the victim's system when the index is loaded from disk. This can lead to full compromise of the affected environment. The vulnerability requires local or limited access to supply the crafted directory and user interaction to load the data. The CVSS 4.0 score is 8.4, reflecting high impact with low attack complexity but requiring user interaction.
Mitigation Recommendations
No official patch or fix is currently documented. Users should avoid loading indexes from untrusted or user-supplied persist directories. Validate or sanitize input directories before deserialization. Monitor vendor advisories for updates or official fixes. Patch status is not yet confirmed — check the vendor advisory for current remediation guidance.
CVE-2024-14021: CWE-502 Deserialization of Untrusted Data in run-llama llama_index
Description
LlamaIndex (run-llama/llama_index) versions up to and including 0.11.6 contain an unsafe deserialization vulnerability in BGEM3Index.load_from_disk() in llama_index/indices/managed/bge_m3/base.py. The function uses pickle.load() to deserialize multi_embed_store.pkl from a user-supplied persist_dir without validation. An attacker who can provide a crafted persist directory containing a malicious pickle file can trigger arbitrary code execution when the victim loads the index from disk.
CVSS v4.0
Score 8.4high
Affected software
Run on your own infrastructure? Check whether these packages are installed with threat-finder — our free open-source scanner.
Weaknesses
AI-Powered Analysis
Machine-generated threat intelligence
Technical Analysis
CVE-2024-14021 is a deserialization of untrusted data vulnerability in the run-llama llama_index project, specifically in the BGEM3Index.load_from_disk() method located in llama_index/indices/managed/bge_m3/base.py. The method uses Python's pickle.load() to deserialize the multi_embed_store.pkl file from a user-controlled persist_dir without validating the input. This allows an attacker who can control the persist_dir contents to craft a malicious pickle file that triggers arbitrary code execution upon loading the index. The vulnerability affects all versions up to and including 0.11.6.
Potential Impact
An attacker who can supply a malicious persist directory to the vulnerable function can achieve arbitrary code execution on the victim's system when the index is loaded from disk. This can lead to full compromise of the affected environment. The vulnerability requires local or limited access to supply the crafted directory and user interaction to load the data. The CVSS 4.0 score is 8.4, reflecting high impact with low attack complexity but requiring user interaction.
Mitigation Recommendations
No official patch or fix is currently documented. Users should avoid loading indexes from untrusted or user-supplied persist directories. Validate or sanitize input directories before deserialization. Monitor vendor advisories for updates or official fixes. Patch status is not yet confirmed — check the vendor advisory for current remediation guidance.
Technical Details
- Data Version
- 5.2
- Assigner Short Name
- VulnCheck
- Date Reserved
- 2026-01-09T20:42:56.495Z
- Cvss Version
- 4.0
- State
- PUBLISHED
Threat ID: 69658281da2266e838450d16
Added to database: 01/12/2026, 23:23:45 UTC
Last enriched: 07/30/2026, 00:29:45 UTC
Last updated: 09/10/2026, 19:36:47 UTC
Views: 356
Community Reviews
0 reviewsCrowdsource mitigation strategies, share intel context, and vote on the most helpful responses. Sign in to add your voice and help keep defenders ahead.
Want to contribute mitigation steps or threat intel context? Sign in or create an account to join the community discussion.
Actions
Updates to AI analysis require Pro Console access. Upgrade inside Console → Billing.
Need more coverage?
Upgrade to Pro Console for AI refresh and higher limits.
For incident response and remediation, OffSeq services can help resolve threats faster.
Latest Threats
Check if your credentials are on the dark web
Instant breach scanning across billions of leaked records. Free tier available.