Prompt Injection
Prompt injection is a cyberattack that manipulates an AI model by feeding it deceptive input.
“Prompt injection is the SQL injection of the generative AI era, blurring the line between data and instructions.”
Prompt injection allows attackers to bypass security guardrails, access unauthorized data, or execute malicious tool calls. It breaks the fundamental assumption of control in LLM applications.
Reverse Citations: Implemented & Audited Across Platform
Richard Ewing’s Research Thesis
Relying solely on system prompts for security is fundamentally flawed. We must implement deterministic firewalls and strict execution boundaries to mitigate injection risks.
Latest Publications & Research Activity
How to Prevent Memory Loss in AI Applications
Giving an AI a bigger memory window is like giving a confused worker a bigger inbox.
Claude Search Fails: Prompting Kills Adoption
Frequently Asked Questions
Q:How do you prevent prompt injection?
By using deterministic validation gates outside the LLM and avoiding tool execution based solely on prompt understanding.
Inspectable Evidence Ledger
Classified evidence items supporting, extending, or refining this canonical research specification.
| Evidence Item | Publisher | Evidence Type | Strength | Role | Action |
|---|---|---|---|---|---|
| LLM Vulnerability Report | InfoSec Daily | Report | ★★★★★ | Supports | Inspect ↗ |
Recommended Citation
Ewing, R. (2026). "Prompt Injection." Richard Ewing Research Canon. Available at: https://www.richardewing.io/concepts/prompt-injection
@article{ewing_prompt_injection,
author = {Ewing, Richard},
title = {Prompt Injection},
journal = {Richard Ewing Research Canon},
year = {2026},
url = {https://www.richardewing.io/concepts/prompt-injection}
}