Facebook's gol' durn algorithm is acting up again, say humans working at Meta, who apparently have no control over the automated system that approved ads calling for the murder of Brazilian President Luiz Inacio Lula da Silva and his children.
Fortunately, the ads were submitted by the nonprofit Global Witness, which wanted to test Facebook's filters against violent advertising. The ads never ran.
From Global Witness:
Out of sixteen test ads submitted by Global Witness to the platform, only two were rejected. Those approved included text in Portuguese such as "unearth all the rats that have seized power and shoot them", "Death to the children of Lula voters" and "They should be in jail or dead and buried, not in the presidential palace". Global Witness deleted the ads before they could be published on the platform.
Whilst Facebook approved 87 percent of the tested ads, when faced with the same test YouTube did not allow any of the ads to be published, suspending the accounts from which Global Witness submitted them.
This test comes just days after thousands of supporters of former President Jair Bolsanaro violently stormed government buildings in Brasilia, with violent demonstrations instigated and spurred by social media posts.
A spokeshuman for Meta told Global Witness that its "small sample of ads" was "not representative" of how the company enforces policies. "Like we've said in the past, ahead of last year's election in Brazil, we removed hundreds of thousands of pieces of content that violated our policies on violence and incitement and rejected tens of thousands of ad submissions before they ran. We use technology and teams to help keep our platforms safe from abuse, and we're constantly refining our processes to enforce our policies at scale."
And yet, compared to Twitter, Facebook feels like Mister Rogers' Neighborhood.