Facebook says group used AI-generated faces to push pro-Trump, anti-Chinese government messages

Facebook has taken down a well-financed campaign that used dozens of artificially generated faces to spread pro-Trump and anti-Chinese government messages, the company and outside researchers said on Friday.

Researchers from New York-based Graphika and the Digital Forensics Research Lab, an arm of the Washington-based Atlantic Council, said it was the first time they had seen the large-scale use of computer-generated faces to spread disinformation on social media.

The researchers said in a report that while telltale signs such as misshapen ears and distorted backgrounds had helped them identify the fakes, “this technology is rapidly evolving toward generating more believable pictures.”

Facebook said 610 Facebook accounts, 89 Pages, 156 Groups and 72 Instagram accounts were involved in the network. The social media giant said those behind the operation had spent upwards of US$9 million on advertising to promote their content, which touched on hot-button issues such as “impeachment, conservative ideology, political candidates, elections, trade, family values and freedom of religion.”

The amount is almost 100 times what Facebook said Russia’s infamous troll farm spent in the run-up to and shortly after the 2016 US presidential election.

Facebook said its investigation linked the activity to US-based Epoch Media Group and individuals in Vietnam working on its behalf, though the accounts more openly associated with The BL, for the Beauty of Life.