Twitter shared data on 5,929 of the more than 88,000 accounts originating in Saudi Arabia that were permanently suspended late last year for violating its platform manipulation policies.
The social network said in a blog post that its investigations traced the coordinated inauthentic activity to Saudi Arabia-based social media marketing and management company Smaat, which appeared to have created, purchased and/or managed the suspended accounts on behalf of—but not necessarily with the knowledge of—its clients.
Twitter said Smaat’s access to the social network and the accounts of its senior executives were permanently suspended, adding that the company managed accounts for high-profile individuals and government departments in Saudi Arabia.
Twitter Safety wrote, “Many of the accounts involved in the overall network employed third-party automated tools in order to amplify non-political content at high volumes. As a general matter, the use of automation to tweet helpful content—like crisis response information or weather updates—is not a violation of our rules. However, this behavior was, in part, strategically employed in an attempt to mask the overall platform manipulation originating from these accounts. These tactics made it more difficult for observers to identify political tweets in the timelines of accounts, which mostly shared automated, non-political content.”
Twitter said content posted by the suspended accounts amplified messages that were favorable to Saudi authorities, using inauthentic engagement tactics such as aggressive liking, retweeting and replying, adding that while most of the content was in Arabic, some was related to events that were relevant to Western audiences, such as sanctions in Iran and Saudi government officials appearing in Western media.
Twitter Safety wrote, “In order to protect the privacy of potentially compromised accounts repurposed to engage in platform manipulation, and in response to researcher feedback requesting that we pre-filter unrelated spam, we have not disclosed data for all 88,000 accounts. In the interest of offering meaningful transparency, the dataset we are disclosing includes a representative, random sample of the fake and spammy accounts associated with this broader network.”
