Facebook and Twitter have removed social media accounts linked to the Russia-backed Internet Research Agency (IRA), which has returned with new efforts to sway Americans in the 2020 presidential election using a fraudulent news website and several fake social media profiles. 

The IRA is a Russian organization known for its massive operation to influence the results of the 2016 election with disinformation tactics. Researchers with Graphika who analyzed its latest activity say it's using new methods to achieve a familiar goal: to persuade voters away from the campaign of Joe Biden and Kamala Harris, similar to its efforts against Hillary Clinton in 2016.

Over the past three years, Facebook has removed about a dozen deceptive campaigns linked to people connected with the IRA, the company states in its latest Coordinated Inauthentic Behavior (CIB) report. Its recent investigation into IRA accounts followed a tip-off from the FBI, which warned Facebook about the IRA network's off-platform activity.

As part of its disinformation efforts, the IRA created a website called peacedata[.]net, which was disguised as an independent news outlet working in both English and Arabic. The network seemed designed to target progressive and left-wing readers in the United States and United Kingdom, with personas masquerading as left-wing reporters and editors, Graphika reports.

The personas on PeaceData each had profile pictures generated by artificial intelligence (AI) and maintained a presence on Facebook, Twitter, and LinkedIn. Researchers note this is the first time IRA accounts were seen using AI-generated avatars, marking "an apparent attempt to appear more convincing."

Unlike the personas' profile photos, however, the people writing PeaceData content were real. The website employed "apparently unwitting individuals, typically novice freelance writers, to write its articles." Between February and August 2020, the website published more than 500 articles in English and more than 200 in Arabic. Some were original works, while others were copied.

To maintain its left-wing image, Peace Data published and shared pieces about race protests across the US, corruption, and suffering caused by capitalism, the researchers report. Articles criticized both right-wing and center-left politicians while endorsing progressive and left-wing policies. The site focused on a small number of political groups including Democratic Socialists, environmentalists, and "disgruntled Democrats in the United States." 

"This shows a consistent focus on a core constituency: left-leaning voters in the United States and United Kingdom who may be dissuaded from supporting the more centrist leadership of the Democratic and Labour parties," Graphika researchers state in a report. The group's activity suggests it was trying to develop a left-wing audience and steer it from the Biden campaign.

This marks the smallest IRA effort exposed to date. Facebook found 13 IRA accounts and two pages on its platform, with a combined following of around 14,000 – one-twentieth of the following an IRA account called "Black Matters U.S." built in 2016. Most of the 14,000 followed the Arabic-language page, while the English-speaking follower count was in the low hundreds.

Twitter has suspended five accounts "for platform manipulation that we can reliably attribute to Russian state actors," the company said in a statement. The accounts were linked to PeaceData, it said; their tweets were low-quality and most received little engagement. Twitter says links to content from the PeaceData site will be blocked from being shared going forward.

"This was an operation in the early stages of audience building, taken down while it was still finding its feet," the researchers say. Its Web articles received few reactions, and its social media posts generally did not spark much engagement. Researchers suggest the operation's greatest success – if it had one – was its ability to hire unwitting writers to create articles for PeaceData.

With the 2020 election only two months away, social media companies are under pressure to detect and block disinformation before it spreads. Earlier this year, Facebook shut down hundreds of Russia-linked pages and accounts related to two disinformation operations. Both Facebook and Twitter have announced new policies to fight the problem, including labeling posts that might contain false or manipulated information or fully removing problematic content.