Is this vulnerability remotely exploitable?

Yes. As described, an attacker can exploit the issue remotely by sending crafted input to any API or user interface that triggers the `_route_using_overviews` function, assuming that their content is incorporated into the prompt construction.

What level of access does an attacker need?

The attacker typically only needs the ability to supply or influence the text that becomes part of the overviews or user query handled by the LLM Prompt Handler. No privileged access to the host or the underlying model is required if the application already exposes a network-accessible interface.

What is the safest general fix pattern?

The most robust fix is to redesign prompt construction so that system and routing instructions are completely separated from user inputs and overviews, using structured roles and explicit delimiting. Avoid interpolating raw user text directly into instruction fields; instead, treat it as data that the model should analyze within a constrained context.

Severity:

HIGH RISK

(7.3)

Prompt Injection (CWE-74) in localGPT LLM Prompt Handler

CVE-2026-5002

Published:

Mar 28, 2026

Status:

Deferred

Summary

Affected Context

Exploit Preconditions

Why This Happens

Potential Impact

General Mitigation Guidance

How Teams Detect Issues Like This

Frequently Asked Questions

Source

NDV

Recent Vulnerabilities

CWE-94

HIGH RISK

HIGH

(8.8)

PHP Code Injection via File Upload (CWE-94) in Aero CMS Admin Post Creation

Remote Code Execution by authenticated users via malicious file upload

May 10, 2026

CWE-94

HIGH RISK

HIGH

(8.8)

Code Injection (CWE-94) via Module Parameters in Evolution CMS

Remote Code Execution by Authenticated Users

May 10, 2026

Vulnerability Type

Category:

Prompt Injection / Injection

CWE:

CWE-74;CWE-707

Impact:

LLM Prompt Manipulation and Potential Data Exposure

Severity:

HIGH RISK

(7.3)

CVSS Version:

3.1

Vulnerability Type

Category:

Prompt Injection / Injection

CWE:

CWE-74;CWE-707

Impact:

LLM Prompt Manipulation and Potential Data Exposure

Severity:

HIGH RISK

(7.3)

CVSS Version:

3.1

Catch This CVE While Code Is Being Written

Detect vulnerable code directly in pull requests, so CVEs are fixed before merge.

Get Started

Ship clean & secure code faster

Start Free Trial

Get Pentest Report

Ship clean & secure code faster

Start Free Trial

Get Pentest Report

Ship clean & secure code faster

Start Free Trial

Get Pentest Report

CodeAnt AI is an offensive and defensive security platform.

Defensive + offensive security.

From first line of code to a pentest
that knows every line of it.

SOC 2 Type II certified | CodeAnt AI is independently audited and compliant with SOC 2 Type II security standards

Ask AI for summary of CodeAnt

Defensive + offensive security.

From first line of code to a pentest
that knows every line of it.

Ask AI for summary of CodeAnt

Defensive + offensive security.

From first line of code to a pentest
that knows every line of it.

Ask AI for summary of CodeAnt

Prompt Injection (CWE-74) in localGPT LLM Prompt Handler

Summary

Affected Context

Exploit Preconditions

Why This Happens

Potential Impact

General Mitigation Guidance

How Teams Detect Issues Like This

Frequently Asked Questions

Source

Recent Vulnerabilities

HIGH RISK

HIGH

(8.8)

PHP Code Injection via File Upload (CWE-94) in Aero CMS Admin Post Creation

HIGH RISK

HIGH

(8.8)

Code Injection (CWE-94) via Module Parameters in Evolution CMS

Vulnerability Type

Vulnerability Type

Vulnerability Type

Catch This CVE While Code Is Being Written

Ship clean & secure code faster

Ship clean & secure code faster

Ship clean & secure code faster

Product

Compare

Company

Developers

Legal

Product

Compare

Company

Developers

Legal

Product

Compare

Company

Developers

Legal