NAIROBI: Social media users in Kenya are being exposed to political disinformation, hate speech and incitement against ethnic communities ahead of elections in August, with rights groups warning the posts are stoking tensions around the vote.
A report by the Mozilla Foundation this month highlighted scores of posts on the TikTok app, including videos that likened a political candidate to Hitler and claimed he would attack other ethnic groups once in power.
The posts are raising tensions in a "highly volatile political landscape" which has previously seen deadly post-election clashes, said the report's author, adding that the platforms are failing to effectively clamp down on abuse.
"Hate speech can be hard to determine ... but a lot of this content is so obviously problematic," said Odanga Madung.
"No media house in Kenya would publish it as they would be shut down."
TikTok said it was working to remove harmful content, and many of the videos Madung flagged have been removed or labelled as sensitive since his study was published.
Kenya goes to the polls on Aug 9. General elections in the East African nation are often highly contentious and polarised, with many traditionally voting along ethnic lines. They have sometimes been marred by intercommunal violence.
An estimated 1,300 people died in clashes following a disputed 2007 election, and some Kenyans fear trouble if the outcome of the upcoming vote is contested.
Madung found more than 130 TikTok videos featuring hate and disinformation over a six-month investigation, which had collectively amassed more than 4 million views.
"The platform is not only allowing this harmful content to be published, but it is also algorithmically amplifying it," he said.
TikTok has assigned a dedicated team for the Kenyan elections and was also rolling out products offering authoritative information about the vote, said Fortune Mgwili-Sibanda, the app's head of public policy in Africa.
"We prohibit and remove election misinformation, promotions of violence, and other violations of our policies and partner with accredited fact-checkers," he said in a statement.
But digital rights groups said efforts by big social media platforms to curb harmful content are inadequate, especially in developing countries.
Their moderators lack knowledge of local languages and dialects while rules do not account for specific cultural and societal contexts, allowing problematic content to spread quickly with potentially serious consequences, they said.
Kenya's two main presidential candidates, Deputy President William Ruto and former Prime Minister Raila Odinga, have expressed faith in the electoral process – but the situation remains precarious, according to political analysts.
While much has changed in the last 15 years, including the emergence of a stronger judiciary, Kenya remains "highly vulnerable" to episodes of pre- and post-election violence, a report by the International Crisis Group said this month.
"That is mainly due to the polarised, ethnically driven and personalist politics that has been a feature of electoral competition in the country for decades," it said.
More than half of Kenya's population of 53 million is under age 35, and about 12 million use social media networks, according to website Datareportal.
A Reuters Institute survey last year found that 75% of Kenyan news consumers found it hard to distinguish between real and fake news online, and that more than 60% said they had come across false information about politics in the previous week.
As the ballot nears, platforms are awash with slick graphics, pictures and videos promoting false and misleading information.
Kenya, often labelled the "Silicon Savannah" due to its thriving tech scene, has also seen the proliferation of "disinformation factories" – armies of social media influencers hired to spread misleading political content, Mozilla found.
"We cannot underestimate the power of these giant social media platforms," said Irungu Houghton, executive director of Amnesty International Kenya.
"Online disinformation, where truth is deliberately distorted to drive particular agendas during elections, is extremely dangerous, and it's very difficult to stop once it's out there."
Officials from the National Cohesion and Integration Commission (NCIC), a government body set up after the 2007 electoral violence to combat hate speech, said they were seeing a clear rise in harmful political rhetoric online.
"A lot of this hate speech is organised," said Danvas Makori, a commissioner at the NCIC.
"We are working with platforms like Facebook to identify individual accounts and groups which are clearly spreading inciteful content and taking legal action – but it's still a very slow process."
Makori said the NCIC had identified 83 cases of online hate speech related to the elections and individuals were being prosecuted.
Disinformation has spread across numerous popular platforms as the elections approach.
On Facebook, a manipulated video falsely suggested the former US president Barack Obama had endorsed presidential candidate Ruto, while Twitter saw a trending hashtag earlier this year that spread unfounded claims against his rival Odinga.
Posts identified by Madung on TikTok included a video with a manipulated image of Ruto dressed in a blood-covered shirt and holding a knife, with a caption alleging he was a murderer. The post had over 500,000 views.
Other TikTok videos used coded language labeling some ethnic communities as "madoadoa", a Kiswahili word meaning "stains", which has been listed as hate speech by Kenyan authorities ahead of the polls.
Facebook and Twitter told the Thomson Reuters Foundation that they were taking aggressive steps to combat disinformation, including hiring expert moderators and partnering with independent fact-checkers.
"We remove misinformation which could lead to imminent violence, as well as highly deceptive manipulated media, like deepfakes, and misinformation which could lead to voter suppression," said Facebook owner Meta Platforms in a statement.
Twitter said it was committed "to protect the health of the electoral conversation".
"We use a combination of technology and human review to identify misleading information on the service, and recently expanded our policy and enforcement to address the rise of misinformation across the world," said a Twitter spokesperson.
But digital rights groups said the platforms' actions were grossly lacking, citing cases from Ethiopia and Myanmar where Facebook is accused of failing to police abuse and playing a key role in spreading hate speech that fuelled ethnic violence.
Taking down harmful content often takes weeks or is not removed at all, and current appeal mechanisms are slow to respond to, or even acknowledge concerns, they said.
When content is in a language other than English, the process can be slower, risking the safety and security of individuals or groups, said campaigners.
"It is clear that moderation practices are inadequate. There is a lack of transparency on how many moderators they have, how many languages, dialects they speak and what knowledge they have," said Bridget Andere, Africa policy analyst at digital rights group Access Now.
"These platforms have the resources, but they are simply not investing enough in content moderation, particularly in the Global South." – Thomson Reuters Foundation