Concern 33 · Working draft · editorial review in progress
Fine-tuning poisoning
Fine-tuning poisoning means using a later round of training to introduce unwanted behavior into an existing model.
Understanding the concern
Fine-tuning adapts a model to a task. That extra training can alter safeguards as well as skills. Research also shows that safety can weaken during some nonmalicious fine-tuning, so weakened safeguards alone do not prove poisoning.
This entry concerns deliberate contamination during that later stage. The distinction matters because prevention must address both malicious data and accidental side effects. A model should be tested again after customization rather than borrowing the original model's safety claims.
An example
A well-trained employee receives a new handbook containing misleading instructions; earlier good training does not guarantee the new instructions will be ignored.
Illustrative unless identified as a documented case.
Words in plain English
- Model
- The learned part of an AI system. It uses patterns found during training to produce an answer or prediction.
- Training
- The process of using examples to shape how an AI model behaves. Learning patterns from material does not establish that the material is true or fair.
- Data poisoning
- Deliberately changing training material to push a system toward unwanted behavior.
- Fine-tuning
- Additional training used to adapt a model for a task or setting. It can change behavior and needs its own checks.
Evidence and its limits
The study examines safety changes under malicious and benign fine-tuning; this entry separates intent from outcome.
Qi and colleagues · Research study
Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To! ↗ (opens in a new tab)Tests safety changes under both malicious and benign customization; accidental weakening is distinct from intentional poisoning.
Broader frameworks and background
These resources provide context. They do not establish every claim in this concern.
A question to discuss
What changed during customization, and were safety checks repeated afterward?
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.
Truthfulness
Truthfulness means making claims that match the evidence and being clear about what we do not know. It includes correcting mistakes rather than protecting a polished story.
Check sources, label guesses, and distinguish a trial result from a proven benefit. If AI helped produce a claim, someone still needs to take responsibility for checking it.
Ask together: What can we honestly say, and what still needs checking?
Justice
Justice means fair treatment, fair procedures, and attention to who receives benefits and burdens. It asks whether people with less power can be heard and seek repair.
Look at outcomes across different groups and individual cases. Provide a usable way to challenge decisions and correct harm. Equal wording in a policy does not guarantee equal treatment in practice.
Ask together: Who benefits, who bears the cost, and who can challenge the decision?
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?
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.