Skip to content
Submit incident
AII-1001-85647
ReportedReported by fewer than three independent publishers, with no official finding.

SQL injection vulnerability in llama_index JSONalyzeQueryEngine

A SQL injection vulnerability was discovered in llama_index's JSONalyzeQueryEngine, which executes LLM-generated SQL queries. The developers moved the component to experimental status and marked it deprecated, recommending against production use without heavy sandboxing.

20 March 2025 · run-llama · Security and cyberattacks
20 March 2025

Drafted by Raidu Incident Detection, an automated system.

AII-1001-856472 sources ↓
Quick facts
Category
Security and cyberattacks
Source
github.com
Date
20 March 2025
Drafted by
Raidu Incident Detection

What happened

A SQL injection vulnerability was identified in the JSONalyzeQueryEngine component of llama_index. The engine executes SQL prompts generated by an LLM with SQLite, potentially allowing arbitrary file creation on the machine running the tool.

Index analysis

The vulnerability was addressed by moving JSONalyzeQueryEngine to the experimental module and marking it as deprecated, with a warning that the tool is not recommended for production use without heavy sandboxing or virtual machines.

Classification

Category
Security and cyberattacks
Harm status
Potential harm
Intent
Unintentional
Timing
Unknown
Failure modes
Prompt injection
AI systems
llama_index (run-llama)

Sources and evidence

This record was researched and written by the Index, based on the following original reporting.

github.com
https://github.com/run-llama/llama_index/commit/bf282074e20e7dafd5e2066137dcd4cd17c3fb9e
huntr.com
https://huntr.com/bounties/095f9e67-311d-494c-99c5-5e61a0adb8f3