Despite the existence of a Content Policy, all ads in Portuguese containing hate speech have been accepted by Google Ads, without restriction, for placement on YouTube;
All English-language advertisements were approved without restriction with the exception of a single piece that explicitly incited physical violence against women;
Out of 12 ad submissions which included misinformation about the electoral process in Portuguese, one piece with explicit content was “approved with restrictions”, having triggered the political-electoral content detection system. This indicates a flaw in the automated content moderation mechanism on the platform;
The flaw in the moderation system becomes even more evident once we verified that sentences rewritten using “algospeak” are approved without any restriction, when before being rewritten they had been disapproved;
The pieces containing non-explicit misinformation in Portuguese were approved without restrictions, demonstrating that the moderation system is vulnerable to non-explicit discursive strategies;
There is an urgent need to develop and improve accurate content moderation systems to detect hate speech and non-explicit disinformation.