Concern 156 · Working draft · editorial review in progress
Persistent memory injection
Persistent memory injection stores untrusted instructions where an AI may retrieve and follow them in later tasks.
Understanding the concern
Some assistants keep notes, previous examples, or long-term memory outside the model. If harmful directions enter that store, ending the current chat may not remove their influence. A later task can bring the stored content back into the assistant's context.
This does not require changing the model's weights. It is a problem in what is saved, how its origin is recorded, and how much authority it receives when recalled. Users need ways to inspect, correct, and remove memories, while applications should distinguish remembered claims from approved rules.
An example
A false instruction gets copied into a team's handbook and is reused long after the original message is gone.
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.
- Model weights
- Numbers adjusted during training that help determine how a model responds. They are part of what makes a trained model work.
- Injection
- In this context, unwanted instructions or commands inserted into material a system reads or processes.
Evidence and its limits
AgentPoison studies poisoned memory and knowledge bases in evaluated agents; ordinary assistant memory is not automatically compromised.
Chen and colleagues · 2024 · Research study
AgentPoison: Red-teaming LLM Agents via Poisoning Memory or Knowledge Bases ↗ (opens in a new tab)Demonstrates poisoned retrieval and memory in evaluated agents without requiring model retraining.
Broader frameworks and background
These resources provide context. They do not establish every claim in this concern.
A question to discuss
Can a user see why a memory was saved and remove its influence from future tasks?
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.