Brand Safety Classifier

Protect your monetization

Classify any video or post by monetization tier before you publish. Avoid yellow flags and demonetization.

Primary use cases

Pre-publish safety check
Advertiser-friendly review
Compliance spot-check

Run this on your prompt →Try the classifier live

Why you can’t just prompt-engineer this yourself

The evolution story

A standard ChatGPT prompt hits 77% on advertiser-friendly classification. Shield was bred through generations of 15 agents each, scored against thousands of examples, to learn the MONETIZABLE / LIMITED / DEMONETIZED boundaries YouTube's own reviewers use. A prompt engineer would need 2-3 weeks to match this.

Live demo

Try Shield — Gen 1.0

0/8000 characters

Free trial: 10 runs across 7 days · No credit card · No signup needed to start

Evolutionary lineage

Shield is a Gen 1.0 agent

1.0

Generation 1.02026-04-14

Initial evolved release. 100% accuracy on 28-sample blind holdout.

Measured accuracy: 100%

This is one of 11 case studies proving the Cambrian evolution engine works. Your prompt can follow the same arc.

Want these numbers on your own prompt?

Bring your production prompt and a labeled dataset. We’ll evolve it the same way we evolved Shield. First evolution is free.

Evolve my prompt →