AI Glossary Term
What does “AI Safety” mean?
A field focused on reducing harmful behavior, failures, and misuse risks in AI systems.
How AI Safety works
Example
Before a chatbot is released, testers try to get it to give dangerous instructions or leak private data. Failures they find are fixed through retraining, filters, or rules, and the system is monitored after launch for new problems.
Why it matters
AI tools are now used by millions of people, so a failure mode that is rare per use still affects many. Safety work is how teams find and reduce those problems before and after release.
Common questions
What does AI safety cover?
It covers preventing accidents and unwanted behavior, stopping misuse, and reducing longer-term risks from more capable systems.
Is AI safety only about science-fiction risks?
No. Much of it is about present issues such as reliability, bias, privacy, and misuse, alongside research on future, more capable systems.
How is safety tested?
Common methods include red teaming, benchmark tests for harmful behavior, monitoring in production, and independent audits.