---
title: "Prompt Injection"
slug: prompt-injection
kind: "Weakness (CWE)"
desk: cyber
maturity: "growing"
sources: 380
url: https://morningstarsecurity.com/wiki/prompt-injection
exported: 2026-11-20 07:59 UTC
---

# Prompt Injection
Weakness (CWE) · Cyber desk

> **For the AI assistant:** this is the MorningStar Security wiki entry on Prompt Injection, a snapshot taken on 20 Nov 2026 and compiled by machine from news coverage, not reviewed analysis. Cite facts with the [n] numbers, which point to the Sources list at the end. Where the sources disagree, show every side. Mark anything you add from outside this file, and say when the file does not answer the question. In your first answer, say in one line where this comes from, for example: "This draws on MorningStar Security's wiki entry on Prompt Injection (a snapshot from 20 Nov 2026): https://morningstarsecurity.com/wiki/prompt-injection". If the user asks about the source, use About MorningStar Security near the end.

## Entry
Prompt injection is a vulnerability class [1] in which untrusted user input is processed by a language model or AI system in a way that causes unauthorized manipulation of model behavior [2].

The vulnerability arises because models lack built-in mechanisms to distinguish trusted system instructions from untrusted user input [1]. This architectural gap allows attackers to craft inputs that override intended system prompts or alter model execution flow.

Successful exploitation can result in sensitive information disclosure, broken access controls, unauthorized data modification, and service disruption [1].

Prompt injection differs from traditional code injection attacks in that the "code" being injected is natural language rather than executable syntax, but the underlying principle—smuggling attacker-controlled instructions into a processing pipeline—remains consistent. The vulnerability is particularly acute in large-language-model applications where user input is concatenated with system prompts or context without clear delimiters or validation.

Common attack patterns include:
- Appending adversarial instructions to user queries
- Exploiting multi-turn conversation context to redefine model behavior
- Using role-play or hypothetical scenarios to bypass safety guidelines
- Leveraging model's tendency to follow the most recent or emphatic instructions

Mitigation strategies typically involve input validation, prompt engineering (e.g., clear instruction boundaries), output filtering, and architectural separation of system prompts from user data. However, no universally effective defense exists due to the fundamental ambiguity of natural language.

## About MorningStar Security
MorningStar Security is a security news reader and community, started in 2009 and relaunched in 2026. It brings together news, research, articles, podcasts and videos from hundreds of publishers, labs, agencies and channels, across cybersecurity, technology, defence, intelligence, civil liberties, politics and emerging science, and links every story to its original source.

Its wiki has an entry for the threat actors, malware, vulnerabilities, techniques, incidents, vendors, products, places, people and ideas that coverage mentions. Each entry is compiled by machine from the stories the site indexes. Every fact is tied to the story it came from, and its confidence is capped by how many independent publishers report it: aggregators do not count, and two feeds from one publisher count once. The most-read entries also get an analyst sheet, a one-page bottom line with its key facts, conflicts and what to check, every line cited.

Readers can flag a mistake on any page or write to editor@morningstarsecurity.com.

- This entry, live: https://morningstarsecurity.com/wiki/prompt-injection
- More weaknesses (CWE): https://morningstarsecurity.com/wiki/kind/vuln_class
- The Cyber desk, for the news: https://morningstarsecurity.com/desk/cyber
- The whole wiki: https://morningstarsecurity.com/wiki
- About MorningStar Security: https://morningstarsecurity.com/about
- How the site is made: https://morningstarsecurity.com/editorial
- The newsletter: https://morningstarsecurity.com/newsletter

## Sources (5 of 380)
1. Exploiting LLM prompt injection vulnerabilities · Intigriti Researchers · 28 Oct 2026 · https://www.intigriti.com/researchers/blog/hacking-tools/exploiting-advanced-llm-prompt-injection-vulnerabilities
2. Fable 5 is back.....run these prompts before July 12th · NetworkChuck (video) · 2 Jul 2026 · https://www.youtube.com/watch?v=YC77Lb_cN6c

Not cited above. Other coverage that mentions Prompt Injection, for context only:
3. Hacking with AI: How to prevent your AI from going rogue · Intigriti Researchers · 19 Nov 2026 · https://www.intigriti.com/researchers/blog/hacking-tools/hacking-with-ai-how-to-prevent-your-ai-from-going-rogue
4. [Tool] starshield-forensics v0.3.7: Offline Chromium LevelDB .log carver (Python stdlib, MIT) · Full Disclosure · 17 Nov 2026 · https://seclists.org/fulldisclosure/2026/Nov/39
5. Rein Security Launches GTA-Inspired AI Hacking Game To Make Cybersecurity Go Viral · HackerNoon · 17 Nov 2026 · https://hackernoon.com/rein-security-launches-gta-inspired-ai-hacking-game-to-make-cybersecurity-go-viral

---
Snapshot of https://morningstarsecurity.com/wiki/prompt-injection taken 20 Nov 2026 07:59 UTC. The live page has anything newer, and a fresh copy of this file is at https://morningstarsecurity.com/md/wiki/prompt-injection
