[ GLOSSARY ]
What are AI guardrails?
Definition
AI conceptsGuardrails
Guardrails are rules and checks that limit what an AI will do, such as refusing certain requests, withholding final answers or pointing a student to help.
Why it matters for schools
Two lessons from the research. First, many guardrails are defaults rather than locks: students can leave consumer study modes, and teacher-configured bots followed intended rules 70.4% of the time. Second, guardrails stop harm but do not create gains on their own: a hint-first AI removed the harm of answer-giving AI but did not beat no AI. Structure and teachers create the gains.
[ FAQ ]
Questions and answers
What are AI guardrails?
Rules and checks that limit what an AI does, such as refusing requests or withholding answers.
Do AI guardrails work?
Partly. They leak, and they prevent harm more than they create learning gains.
Which guardrails matter most for schools?
Learning rules students cannot switch off, a human safety route, and limits on data use.
[ SOURCES ]
Sources
Facts checked Oct 18, 2026.
- 1Riahi et al., arXiv preprint on teacher-configured chatbotsSep 24, 2026 · arxiv.org
- 2Bastani et al., field experiment on generative AI in high-school math, PNASJun 25, 2025 · doi.org
- 3Common Sense Media, ChatGPT for Teens risk assessmentOct 7, 2026 · institute.commonsensemedia.org
[ AI FOR SCHOOLS ]
The AI your school can approve.