Refusal behavior
Refusal behavior is a model’s response pattern of declining to fulfill a request or perform a requested task.
LLM Customization and Fine-Tuning
A model's act of declining to fulfill a request, often because complying would be inappropriate or unsafe.
- How it is used
- Protects against unsafe compliance when models are distilled or adapted for enterprise tasks.
- How deep it goes
- Explains failure mechanisms and gives data, testing, and mitigation steps, including deployment thresholds.
- Examples
- Uses an IT password request and a four-prompt safety check that includes a phishing request.
- What you can do after
- Diagnoses safety regressions by comparing models and applying refusal-data, fine-tuning, or filtering mitigations.
Read how LLM Customization and Fine-Tuning explains refusal behavior →
Build an AI Agent (From Scratch)
A model refusal is a model response that declines to answer or perform a requested task instead of attempting it.
- How it is used
- Guides production-agent design by treating refusals as model-dependent behavior requiring testing and fallback handling.
- How deep it goes
- Defines refusals, illustrates obfuscation-triggered behavior, and recommends empirical testing, but does not specify undocumented trigger rules.
- Examples
- Uses reversed text asking for the opposite of “left,” plus refund, account-deletion, and financial-advice boundaries.
- What you can do after
- Operate/configure production agents by testing target models, checking use policies, and preparing fallback strategies for refusals.
Read how Build an AI Agent (From Scratch) explains refusal behavior →
What to know first
- AI agentAn AI agent is a computational system that uses artificial intelligence to interpret information and select or perform actions toward goals while interacting with an environment.
Related concepts
- fine-tuningproduced by · Fine-tuning is the continued training of a pretrained model on targeted data or objectives to adapt its parameters or behavior to a task, domain, or use.
- safety regression suitemeasured by · A collection of tests for detecting safety failures introduced by model changes.
- behaviorkind of · Behavior is the observable actions, responses, outputs, or effects of a system or object under specified conditions or interactions.
- adversarial inputused by · An input crafted, selected, or modified to exploit a system’s weaknesses or induce incorrect, unsafe, or otherwise unintended behavior.
- preference dataproduces · Preference data is a collection of judgments about the relative preference or quality of alternatives, often candidate outputs for the same input.
Questions about refusal behavior
What is refusal behavior?
Refusal behavior is a model’s response pattern of declining to fulfill a request or perform a requested task.
Which Manning books explain refusal behavior?
LLM Customization and Fine-Tuning and Build an AI Agent (From Scratch).
What should I know before learning refusal behavior?
Manning books that teach refusal behavior usually introduce AI agent first.
What is refusal behavior a kind of?
Refusal behavior is a kind of behavior. Behavior is the observable actions, responses, outputs, or effects of a system or object under specified conditions or interactions.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.

