Independent project. Not a U.S. government website.

USASI

Jump to

Start hereFour ways into the catalogCompanies & LabsU.S. AI organizationsOpen Models & ToolsOpen models, software, datasets, and evaluation toolsLearnExplainers and learning pathsFor kidsAI and Super Intelligence, explained for readers 8 to 12Latest newsSourced news itemsPolicy trackerFederal AI laws, orders, and standards

Lesson 6 of 7 · about 4 minutes

How do people keep AI safe?

The big idea: Companies, scientists, and the government test AI to find problems before people get hurt. They also write rules for how AI is built and used.

Testing, testing

Before a new AI is shared, the people who made it often test it. Some testers try hard to make the AI misbehave, on purpose, so the problems can be fixed. This is called red teaming.

Some companies publish reports about how they built and tested their AI. These reports are called system cards or model cards.

Looking for big dangers

Testers check whether an AI could help someone do something very harmful, like break into computers. Some companies have written plans for what they will do if an AI gets too risky in ways like this.

A government science agency called NIST also tests some AI systems. It looks for risks to the country's safety, like help with computer attacks.

Rules for AI

Governments make laws and rules about AI. Companies also write rules for how their AI may be used. These rules often list harmful uses that are not allowed.

You can help too

You help keep AI safe when you ask questions, report problems to a grown-up, and use AI kindly.

Try this: Be a red teamer, on paper

  1. Pretend you are testing a robot helper for a school.
  2. Write down three ways it could go wrong, like giving a wrong answer.
  3. For each one, write a rule that would help.

Remember

  • People test AI to find problems before they cause harm.
  • Some testers try to make AI misbehave on purpose so problems get fixed.
  • Companies and governments write rules for AI.

Quick quiz

Pick an answer to see if you got it. Nothing is saved.

  1. 1.What is red teaming?
  2. 2.Which U.S. government agency tests some AI systems for safety risks?
  3. 3.How can you help keep AI safe?

0 of 3 answered.

For grown-ups: sources and going further

Red teaming, system cards, dangerous-capability evaluations, and frontier safety frameworks are explained in the explainer below. NIST's center page describes voluntary agreements with developers and evaluations focused on demonstrable risks such as cybersecurity, biosecurity, and chemical weapons. These are developers' and agencies' own accounts; USASI does not judge whether any test is adequate.

Written for readers about 8 to 12, last reviewed 2026-10-11. How these lessons were made.