YouTube ads hate speech blocklist is 'inconsistent' and barely applied
YouTube ads hate speech blocklist is 'inconsistent' and barely applied
Occurred: April 2021
Page published: April 2021
An investigation revealed that Google’s automated "blocklist" for YouTube ad placements was easily bypassed by hate groups while simultaneously penalising social justice content, causing financial harm to marginalised creators and brand safety risks for advertisers.
According to The Markup, Google’s YouTube blocklist of hate terms is inconsistent and highly permeable, allowing advertisers to target people in terms such as 'white lives matter' and 'white power' but blocking them from running ads against terms such as 'Black Lives Matter.
It also found that less than a third of the 86 hate-related terms tested were blocked.
Google later said it had blocked additional terms associated with hate speech from being used as ad keywords on YouTube videos. The company said it does not publicly state how it develops its enforcement tools so that bad actors are less likely to game its system.
The episode raised questions about Google's commitment to user safety, and the extent to which it permitted the monetisation of hate content to boost its own profits.
Operator: Alphabet/Google/YouTube
Developer: Alphabet/Google/YouTube
Country: USA
Sector: Technology
Purpose: Identify and block offensive ads
Technology: Advertising management system
Issue: Accuracy/reliability; Bias/discrimination; Safety; Transparency
AIAAIC Repository ID: AIAAIC0614