Skip to main content

Data poisoning, backdoors and adversarial examples

Explain how attackers corrupt a model through its training data, fine-tuning data or knowledge base, or fool it with crafted inputs, and put controls at each step of the pipeline.

Pro 1 · Quality engineerabout 3 hours5 sections

Quick answer

Data poisoning plants bad data where a model learns, to make it wrong or hide a backdoor. Evasion feeds a finished model crafted inputs that fool it.

Why it matters

Models learn whatever their data teaches, and most teams train on data they did not write: the web, vendors, user feedback, shared documents. If you cannot say where your data came from, you cannot say what your model will do. Security reviews of AI systems now ask this.

Sign up free to read the whole lesson

The rest of this lesson: 5 parts, video lessons, the best course, practice and a quick check. It is free, takes a few seconds and needs no card.

Accounts are for learners aged 18 and over. Under 18? The quick answer and the first part of every lesson stay open: browse lessons by topic.

On the roadmap: AI Security Engineer

SecurityPro 1: Quality engineerVideo lessonsData poisoningBackdoor attackAdversarial exampleMITRE ATLASFine-tuningRetrieval-augmented generation

Track

Today's AI turns are used

Lessons, roadmaps, videos, practice cards and the code playground stay free with your account, with no limit.