ITG NowHome › AI glossary › Red teaming
What is Red teaming?
Red teaming is when a group plays the attacker on purpose to find weak spots in a system before real attackers do.
- Category Safety & law
- Also called red team
- Updated 20 Oct 2026
Red teaming, explained simply
A red team is a group that acts like an enemy. Their job is to break things so the owners can fix them.
In AI, red teams try to make a model say harmful things or leak data. They record what works.
The makers then fix those gaps before release. Many AI companies do this, sometimes with outside experts.
On ITG Now, our daily AI news stories sometimes mention red teaming when a company tests a new model for risks.
Related terms
AI safetyJailbreakPrompt injectionGuardrailsAI alignmentVulnerability
Questions people ask about Red teaming
Why do AI companies use red teams?
It is safer to find problems in testing than after launch. Red teams think like attackers to spot them early.
Is red teaming the same as hacking?
It uses similar skills but has permission and a goal of fixing problems. Real attackers have no permission.
Can anyone do red teaming?
Companies hire trained experts and also run reward programs. Always follow a program's rules before testing anything.