AI Safety Terms Every Parent and School Should Know

The expectation that student work honestly reflects their own learning.

A written policy outlining how students and staff may use technology — including AI.

An AI system that can take actions on its own — browsing, booking, emailing — not just generating text.

The field of making AI systems behave in line with human values and intent.

The ability to understand, evaluate, and responsibly use AI tools.

Using AI as a learning partner — for explanation, practice, feedback, or differentiation.

Monitoring how a user interacts with a product over time to build a profile or personalize content.

Data derived from physical or behavioral traits — face scans, voiceprints, typing patterns.

The Coalition for Content Provenance and Authenticity — an open standard for cryptographically signing media origin.

AI tools used directly in instruction — tutors, graders, lesson planners, and assistants.

AI products designed to simulate friendship, romance, or emotional support.

A clear, informed agreement to a specific use of personal data.

The systems — automated and human — that filter unsafe content from an AI product.

The maximum amount of text an LLM can consider at once.

The U.S. Children's Online Privacy Protection Act, which regulates how online services collect data from children under 13.

Deceptive UX choices that push users toward decisions they would not otherwise make.

The principle of collecting only the data strictly necessary to deliver a service.

The length of time a service keeps user data after it is no longer actively needed.

AI-generated media that makes a real person appear to say or do something they never actually said or did.

Educational technology — software, devices, and platforms used in learning environments.

When a user — especially a child — develops reliance on an AI for comfort, validation, or companionship.

The U.S. Family Educational Rights and Privacy Act, which protects the privacy of student education records.

Customizing a pre-trained AI model on a narrower dataset to specialize its behavior.

The European Union's General Data Protection Regulation, the world's strictest general data privacy law.

The child-specific provisions of GDPR governing minors' personal data.

AI systems that create new content — text, images, audio, video, or code — based on a prompt.

Built-in safety rules that limit what an AI model will say, do, or generate.

When an AI confidently states something that is false or fabricated.

The act of an AI model generating an output from an input — i.e., running the model.

A crafted prompt designed to trick an AI into bypassing its safety rules.

Large Language Model — an AI trained on massive amounts of text to predict and generate language.

AI that handles multiple input types — text, image, audio, video — in one model.

A one-sided emotional bond a user forms with a media figure or AI persona.

Product patterns engineered to maximize attention, engagement, or emotional response.

The instruction or question a user gives to an AI model.

Verifiable information about where a piece of content came from and how it was made.

When a service sends user data to outside companies — advertisers, analytics vendors, or other AI providers.

A small chunk of text (often a few characters) that an LLM processes one at a time.

The dataset used to teach an AI model.