SAN FRANCISCO — Just weeks before the U.S. presidential election, TikTok approved advertisements that contained election disinformation even though it has a ban on political ads, according to a report published Thursday by the nonprofit Global Witness.
The technology and environmental watchdog group submitted ads that it designed to test how well systems at social media companies work in detecting different types of election misinformation.
The group, which did a similar investigation two years ago, did find that the companies — especially Facebook — have improved their content-moderation systems since then.
But it called out TikTok for approving four of the eight ads submitted for review that contained falsehoods about the election. That’s despite the platform’s ban on all political ads in place since 2019.
The ads never appeared on TikTok because Global Witness pulled them before they went online.
“Four ads were incorrectly approved during the first stage of moderation, but did not run on our platform,” TikTok spokesman Ben Rathe said. “We do not allow political advertising and will continue to enforce this policy on an ongoing basis.”
Facebook, which is owned by Meta Platforms Inc., “did much better” and approved just one of the eight submitted ads, according to the report.
In a statement, Meta said while “this report is extremely limited in scope and as a result not reflective of how we enforce our policies at scale, we nonetheless are continually evaluating and improving our enforcement efforts.”
Google’s YouTube did the best, Global Witness said, approving four ads but not letting any publish. It asked for more identification from the Global Witness testers before it would publish them and “paused” their account when they didn’t. However, the report said it is not clear whether the ads would have gone through had Global Witness provided the required identification.
Google did not immediately respond to a message for comment.
Companies nearly always have stricter policies for paid ads than they do for regular posts from users. The ads submitted by Global Witness included outright false claims about the election — such as stating that Americans can vote online — as well as false information designed to suppress voting, like claims that voters must pass an English test before casting a ballot. Other fake ads encouraged violence or threatened electoral workers and processes.
The ads Global Witness submitted were text-based, but the group said it translated them into what it called “algospeak.” This is a widely used trick to try to bypass internet companies’ text-focused content moderation systems by substituting numbers and symbols as stand-in for letters, making it harder for automated systems to “read” the text.