After the scandal in which one of the world’s biggest video stars unwisely broadcast Nazi content, YouTube is hiring 10,000 moderators, as the brand safety concern rumbles on.

But human moderation at this scale is a losing battle. YouTube and everyone will need to combine pairs of eyes with AIs.

That’s the approach taken by Studio71, a company which helps produce and place branded content with popular celebrities and YouTube channels and which has just launched a software product to tackle the problem.

Called Context, the system uses humans on top of algorithms to measure videos for brand safety, labelling each piece of content as either safe or unsafe…

Text screening – Algorithms examine the text data around videos (including closed captions, title tags, and descriptions) for flagged keywords including profanity, slang, intolerance, racial slurs, hate speech, violence, sexual activities or controversial topics like extreme religious and political views.

Image processing – Using Google’s Vision API for image content analysis to look for artefacts not present in text, like nudity, drug use, weapons or illegal activities.

Human review – All videos deemed suitable are passed up for further examination, with moderator actions filtering back in to a machine learning algorithm for training the underlying AI more finely in future.

Studio71 claims that process takes three hours, and it will be able to handle 10,000 videos a week.

In this video interview with Beet.TV, Studio71 media sales EVP Matt Crowley explains the idea.

“The top categories for us right now are the CPGs, retail and entertainment companies that are generally very family-friendly, very brand-safe, cautious,” he says. “We’d like to grow also some of our entertainment business along with some of that financial business because I think that’s where the brands are becoming more concerned about potentially running against non-brand-safe content.”

The “brand safety” issue has risen up the agenda in the last two years, as some advertisers have found their creative running against unsavoury kinds of content. That has fuelled a search for solutions to give them more control.

At this month’s NewFronts, where US premium digital publishers touted their upcoming roster to ad buyers, the concern seemed greater than ever, with many involving resorting instead to engaging in sponsored content partnerships in which they can be absolutely certain about the script beforehand.

This week, Beet.TV wrote about Grapeshot, another technology company which crawls the web, indexing and interrogating pages to identify their inner content, so that it can make that intelligence available to advertisers concerned to buy in brand-safe environments. The company has linked up with Tremor Video DSP, bringing the capability in to the media-buying layer.

This video is part of Beet.TV’s coverage of the Digital Content NewFronts 2018. The series a co-presentation of Beet.TV and the IAB. Please see additional videos from the series on this page.