What is Data Poisoning

Definition

Data Poisoning is an attack or bug in which harmful data gets into the training and degrades the behavior of the model. Simply put, this concept helps assess risk, liability, safety, and compliance. In practice, it helps to understand what capabilities the tool actually has, what data it will need, and what limitations are worth checking before implementation.

Example

The attacker adds specially distorted examples to the training set so that the model makes mistakes in the right cases.

Why it matters

Data poisoning is important for AI safety, especially when the model learns from user content. This helps you choose AI tools not by big promises, but by how they work in a real problem.

How it works

First, stakeholders, data, and potential harm are identified, then checks, restrictions, audits, and responsibilities are introduced. In the case of the term “Data Poisoning,” it is important to look at the data, quality criteria, and application conditions separately.

Where it is used

• Important in products where AI impacts people, personal data, security, legal risks or decision making.

Limitations

Risks change as laws, products and data change, so these pages require regular editorial review.