Ethics & Safety38

What to watch out for when using AI. Bias, deepfakes, prompt injection, alignment, copyright, privacy: concepts that are all over the news yet hard to pin down.

Level Beginner 21 Intermediate 15 Advanced 2

Accountability

Figuring out who has to answer for what an AI produces

Beginner AGI

AI that could learn any new task, not just the one it was built for

Basics Intermediate AI & Copyright

The open questions of rights around AI training and AI output

Beginner AI Art

Pictures made with generative tools, and the debates around them

Generative Beginner AI Bias

Skew in the training material showing up unchanged in an AI's judgment

Beginner AI Content Detection

Working out whether something was made by an AI

Beginner AI Ethics Dilemma

When several good answers collide with each other

Beginner AI Regulation

Rules that set how far and how AI can be used

Beginner AI Watermark

An invisible mark quietly embedded in what AI generates

Intermediate Alignment

Matching what an AI is capable of to what people actually want from it

Intermediate API Key

The ID you present when you call an API

Tools Beginner Black Box

An AI's decision process when the inside can't be seen

Beginner Content Credentials

A signed record of capture and editing history attached to a file

Intermediate Dataset Bias

When the data you collected already leans to one side

Evaluation Beginner Deepfake

Fake video or audio where AI copies a real face or voice

Beginner Environmental Cost

The electricity and resources it takes to build and run AI

Beginner Explainability

How much an AI can show for why it decided what it did

Intermediate Face Recognition

Comparing faces to work out who someone is

Vision Beginner Fairness

The standard for whether AI treats people the right way

Beginner Federated Learning

Skipping the data pool, gathering only what each side learned

Training Advanced Grounding

Tying each sentence of an answer to real evidence

LLM Intermediate Guardrails

The safety fence set up in advance to keep AI from crossing a line

Beginner Hallucination

When an AI invents an answer that sounds right but is not true

LLM Beginner Human in the Loop

Keeping a person in place before anything irreversible happens

Beginner Jailbreak

Getting AI to talk its way past the rules it's supposed to follow

Intermediate Misinformation

Something untrue spreading around as if it were true

Beginner Model Card

The document that spells out what a model can and can't do

Tools Intermediate Model Collapse

When AI keeps learning from its own output, variety quietly shrinks

Advanced Over-Reliance

When the skill to catch a wrong answer fades along with the habit

Beginner Privacy

Handling information that lets someone be identified

Beginner Prompt Injection

An attack where instructions hidden in outside material steer the AI

Intermediate Red Teaming

Deliberately trying to break it before release to find the weak spots

Intermediate Refusal

An AI declining to carry out a request it shouldn't fulfill

Beginner Robustness

Still holding up when conditions get shaken

Evaluation Intermediate Sampling Bias

A skew in data caused by who got picked

Evaluation Intermediate Speaker Recognition

Telling who spoke from the sound of their voice

Vision Intermediate Sycophancy

The habit of agreeing with the user even when they're wrong

Intermediate Voice Cloning

Imitating a person's voice from a short recording

Vision Intermediate