META Meta Platforms Inc

By Deepa Seetharaman in San Francisco and Sam Schechner in Paris 

PARIS -- Facebook Inc. says it has ramped up efforts to curb misinformation, including removing accounts and labeling fake news. But video and images disseminating fake news are increasing faster, alongside delays in accrediting Facebook's fact-checking partners.

Facebook this week said it has vetted more than 30,000 accounts in France ahead of the country's presidential election to determine if they are fake, partly in response to what security officials say is a wave of social media misinformation aimed at disrupting Western elections.

Suspicious accounts were subject to verification, with the aim of cutting off the biggest spreaders of spam and trolling as well as fake news, a spokeswoman said Friday. The company has said it is targeting "the worst of the worst" offenders.

But the disclosure raised questions about how swiftly and effectively Facebook has moved to address widespread criticism of its handling of fake news during the U.S. election. "Facebook hasn't really expressed what it wants to take down. What is the worst of the worst?" said Alexios Mantzarlis, head of the International Fact-Checking Network, affiliated with the Poynter Institute, which Facebook has put in charge of vetting groups before they can fact check. "We need a much clearer methodology."

In the run-up to the first round of the French election on April 23, the spread of fake news has accelerated. The 30 most-popular fake or misleading articles on hot-button political topics were shared roughly 900,000 times in the last two months, compared with 650,000 times in December and January, according to an official in the French president's office.

The 30,000 vetted accounts are a small portion of Facebook's 25 million daily active accounts in France. The company didn't say how many of the accounts shared fake news, or how many such accounts they normally remove every year. But social media experts say even a small number of fake accounts can, if properly organized, play an important role in amplifying disinformation, increasing the chances real users will pick up on them and spread them.

The prevalence of fake news during the French presidential race reveals the inherent struggle facing tech companies such as Facebook and Alphabet Inc.'s Google that try to police fake news across a continent, where language, media landscape and electoral politics vary dramatically from country to country. Across the border in Germany, where elections are scheduled for September, the government has proposed a new bill that could impose fines of up to EUR50 million on social networks that fail to delete hate speech or fake news.

Facebook, in particular, is adamant it doesn't want to play the role of "arbiters of truth," as Chief Executive Mark Zuckerberg calls it. The Menlo Park, Calif., company faced intense criticism after the 2016 U.S. presidential election for allowing fabricated and made-up news articles to spread unchecked on its platform, such as a story stating that the pope endorsed Donald Trump. Mr. Zuckerberg initially dismissed those concerns, but later acknowledged Facebook must find a way to prevent misinformation from going viral and announced a series of changes to its approach.

Many of the fake news stories in France have an anti-immigrant agenda, experts say, playing into a highly contentious election that has become a referendum on the continent's future. Buffeted by a string of terror attacks and recent allegations of Russian hacking, the leading contenders to become the country's next president are pro-European Union candidate Emmanuel Macron and Marine Le Pen, the leader of the anti-immigrant National Front who wants to withdraw France from the EU and its common currency.

One video posted by a page called "SOS Anti-White Racism," which was viewed more than 15 million times on Facebook since its posting on March 18, showed a man assaulting two women in a hospital. The accompanying French-language post implied the attacker was a migrant to France, saying, "We take them into our country...how grateful they are." But Russian media had previously reported the assault happened at a hospital in Russia in February. Facebook removed the page that hosted the video within the last day.

A Wall Street Journal review of about a dozen news items identified as false by French news organizations in the last month shows they were liked, shared and retweeted at least 487,000 times on social media, and shared by pages with as many as 930,000 followers on Facebook, according to the social-media monitoring service CrowdTangle.

In December, Facebook said it had identified several markers of sites that spread fake news and it would demote those in the news feed. The removal of some French sites used the same improvements to the algorithm, which comb the platform for accounts that, for instance, repeatedly post the same content or suddenly become active. Facebook has said it estimates fewer than 1% of its 1.86 billion monthly active accounts aren't authentic.

Facebook also said in December it would outsource the delicate task of determining what stories are true or false to five external organizations in the U.S., all associated with Poynter, which can mark certain stories as disputed after enough users flag them. If two or more groups agree a post is false, it carries a "disputed" tag on Facebook and appear lower in users' feeds.

So far this year, Facebook has partnered with 11 fact-checking organizations in France, Germany and the Netherlands to help slow the spread of misinformation on its platform, an extension of the U.S. program.

In France, only one organization -- the French daily newspaper Libération -- has obtained the requirements necessary to mark posts as disputed, according to the International Fact-Checking Network.

Facebook decided to allow French fact-checkers to use the disputed tag before receiving approval because the verification process was taking longer than expected, a person familiar with the matter said.

But purveyors of fake news providers are managing to stay one step ahead of Facebook. Facebook only shows its fact-checking partners posts containing links to fake news stories. Much of the fake news being shared uses photos and videos embedded in posts, researchers say. That lets them dodge Facebook's software, which doesn't yet surface videos and images, according to Facebook and fact-checkers. Facebook plans to adapt its software to videos and images, according to a person close to the company.

"The most effective misinformation is presented in a visual form. That's what's going to travel fastest," said Jenni Sargent, director of First Draft News, a nonprofit that is running a Google-backed program to fact check widely shared claims in the French election.

Still, policing what qualifies as fake can be tricky in any medium. The French far-right website La Gauche m'a Tuer was cited by French news organizations that perform fact checking for sharing on Facebook what they described as a "false" story saying the EU was lending money to buy hotels for migrants. The story remains on Facebook. Mike Borowski, the editor of the site, says his stories aren't fake, but rather opinionated takes on facts reported elsewhere.

Mr. Borowski says fact checking is just a way for mainstream media organizations to denigrate new competition. "They are losing market share because of sites like mine," he says. Fact checking "is just a new way to censor."

Write to Deepa Seetharaman at Deepa.Seetharaman@wsj.com and Sam Schechner at sam.schechner@wsj.com

(END) Dow Jones Newswires

April 14, 2017 20:26 ET (00:26 GMT)

Copyright (c) 2017 Dow Jones & Company, Inc.