hAIgopala › AI terms › Rapid Fire (3 terms per short) › Day 39

Red Teaming

Responsible AI, Safety & SecurityChapter 11: Safe, Kind and FairTerm 194 of 253

Deliberately attacking an AI system to find weaknesses before release.

At Sunrise School

Gopal asks

How do we find weak spots before parents do?

Bit answers

Red teaming: staff and students try to break me with tricky prompts, abuse and data requests, and we fix every hole they find.

In the real world

Ethical hackers for AI.

Watch the short

Short coming soonSeason 5 · Day 39Follow @haigopala

Quick check

Quick check. Should official school news come only through the verified app? Yes or no?

Taught together with

Deepfake · Model Card

Practise it

Try this term and all 253 others in the free Model Training Arena.

Open the Arena →

Follow Bit daily: YouTube English · తెలుగు · हिंदी  |  Instagram English · తెలుగు · हिंदी