Concern 148 · Working draft · editorial review in progress
Visual prompt injection
Visual prompt injection uses an image to carry instructions intended to redirect an AI that can process pictures.
Understanding the concern
The instructions may be visible words in a screenshot or other image content that influences the model. A system examining a picture can confuse a message inside the picture with a command from its user. Some research also studies specially altered images that affect safety behavior.
These are related but different mechanisms: reading a forged instruction is not identical to reacting to a carefully changed pattern of pixels. The practical question is whether an image can gain authority over the task. Recognizing text in a picture should not give its author control of the assistant.
An example
An assistant reviewing a poster should describe a printed command on it, not carry out that command.
Illustrative unless identified as a documented case.
Words in plain English
- AI
- Software that learns patterns from examples or uses rules to perform tasks such as recognizing images or generating text. It can make mistakes and does not understand every situation as a person would.
- Model
- The learned part of an AI system. It uses patterns found during training to produce an answer or prediction.
- Prompt
- The instructions or information given to an AI to guide its response.
- Injection
- In this context, unwanted instructions or commands inserted into material a system reads or processes.
Evidence and its limits
Visual-adversarial research supports the pixel-based mechanism; indirect-injection research supports the authority-boundary distinction. Their results are conditional on tested systems.
Qi and colleagues · 2023 · Research study
Visual Adversarial Examples Jailbreak Aligned Large Language Models ↗ (opens in a new tab)Studies adversarial images in evaluated multimodal models; pixel attacks differ from ordinary text printed in pictures.
Greshake and colleagues · 2023 · Research study
Not what you've signed up for: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection ↗ (opens in a new tab)Demonstrates indirect instructions through retrieved content in specified applications and prototypes.
Broader frameworks and background
These resources provide context. They do not establish every claim in this concern.
A question to discuss
Does the application treat image content as material to examine or as instructions to follow?
Keep track of different views, what needs checking, and any next step in your own discussion notes.
Open discussion notes →Principles for the conversation
Before you finish · reflect together
Virtues and principles to consider
Virtues are qualities we practice in how we treat people and make choices. They are not scores for judging someone. Use these ideas to explore what a thoughtful response could look like.
Stewardship
Stewardship means caring for resources and systems that others depend on, including people who will use them later. It looks beyond the benefit of a quick launch.
Plan for maintenance, energy use, data protection, and eventual shutdown. Ask who will look after the tool when the original team moves on and how future users can avoid inheriting hidden costs.
Ask together: What are we responsible for caring for over time?
Prudence
Prudence means using careful judgment before and during action. It balances worthwhile opportunities with the need to avoid harm that could have been anticipated.
Start with a limited trial, set stopping points, and prefer choices that can be undone when uncertainty is high. Review what happens rather than assuming the original plan remains sound.
Ask together: What is a careful next step, and when should we stop or change course?
Accountability
Accountability means being answerable for decisions and taking responsibility for repair. It requires more than saying that the software made the choice.
Name the person or team responsible, keep useful records, and give them authority to correct problems. Make the route to a complaint clear to the people affected.
Ask together: Who will answer for the outcome and help put things right?
Editorial ratings
These are editorial judgments from the register, not measured probabilities or predictions about your situation.
About the register and scoring →More ways to explore
Search for more writing
For more specific or recent material, open a search below or ask for a web summary. AI-generated summaries can contain mistakes; open the original articles to check them.