AI Safety Basics
Learn hallucinations, privacy, bias, verification, and when AI should not make decisions alone.
AI can sound confident and be wrong
A hallucination is when an AI produces information that looks plausible but is false, incomplete, or unsupported. Confidence in the wording is not proof.
Protect private information
Avoid pasting passwords, private documents, private messages, medical details, legal details, school records, or personal data unless you understand the tool, the account settings, and how your data may be used.
Bias can enter through data
AI systems learn from data created by people and organizations. If that data contains bias, missing context, or unfair patterns, the model can repeat those problems in its output. This is why AI answers should be reviewed, especially when people are affected.
High-stakes decisions need humans
Do not let AI make final decisions about health, legal issues, finance, education access, employment, housing, insurance, safety, or other sensitive areas. AI can help prepare questions or organize information, but qualified people and official sources should guide the final decision.
Build a verification habit
For important work, ask AI what might be uncertain, check original sources, compare more than one source, and look for dates, names, numbers, and missing assumptions. Treat AI as a draft partner, not as the authority.
Example
Verification habit
Ask an AI for the latest scholarship deadline and submit without checking the official website.
Ask the AI to create a checklist of what to verify, then confirm the date on the official scholarship page.
Takeaway AI can help you know what to check, but official sources should decide important facts.
Mini challenge
Add a safety check
Sample answer: List which parts of a response might be uncertain and what to verify with an official source.
Open this mission in the Kiwippy app to mark it complete and track progress.