Researchers say Meta is failing to moderate dangerous content on Facebook in Kenya, citing failures in AI as well as linguistic and cultural expertise gaps

NAIROBI — The shooter approaches from behind, raising a pistol to his victim's head. He pulls the trigger and “pop,” a lifeless body slumps forward.

Washington Post 2022-08-01 Neha Wadekar

Context & Ripple Effects

This lands weeks before Kenya's election, after human rights groups found Facebook had approved English and Swahili ads instigating ethnic violence despite Meta's pledge to keep the vote safe. It also extends a documented enforcement gap: sources reported years earlier that third-party moderators were never even instructed to enforce Facebook's own call to arms policy designed to prevent armed intimidation.

First-order effects

Kenyan users are exposed to graphic and inflammatory content through an election cycle Meta publicly committed to securing, while its moderation stack lacks the linguistic and cultural expertise to catch violations in local languages.

Second-order effects

Legal exposure follows the moderation gap into Meta's own courts of convenience: months later, the son of a murdered Ethiopian professor files a lawsuit against Meta in Kenya alleging its algorithms amplify hateful and violent content for engagement.

Third-order effects

The same structural weakness recurs at scale elsewhere — Meta's tools proved unable to parse Arabic dialects during the Israel-Hamas war — and as Meta leans harder on AI moderation, the trade-off sharpens: under-detection in non-English markets coexists with wrongful account deletions that users report on Instagram and Facebook.

The trend: Platform moderation is bifurcating by language: Meta is automating enforcement globally while the linguistic and cultural expertise needed outside English-speaking markets lags, leaving Global South users bearing both missed harm and mistaken takedowns.

Discussion

@moustafaayad @moustafaayad on x
2. We tracked these networks for two years ahead of these polls, and found 30 pages, with just shy of 40,000 followers, 445 accounts, and one group, and analyzed 600 posts in Swahili, Somali and Arabic. We found Swahili calls for violence were the most prevalent.
@iamwalkert Walker Tisdale on x
FYI. We Must Informed of the Policies & Practices of Social Media Platforms We Support.... https://www.washingtonpost.com/ ... #BlackTwitter #BlackTechTwitter #TwitterFriends
@qulshtm Abdullahi Halakhe on x
It is painful to watch how Kenya's election reporting, especially from international media, is pretty much stuck in 2007. Social media's power/leverage is not as huge as reckoned. Mainstream media, especially community radio, wield far more influence. https://www.washingtonpost.c…
@inteltweet Evan Kohlmann on x
The Washington Post: “Why dangerous content thrives on Facebook and TikTok in Kenya” https://www.washingtonpost.com/ ...
@moustafaayad @moustafaayad on x
5. We consistently found posts that were years old, the oldest being an attack on Kenyan soldiers that had been on the platform for more than half a decade. There are real consequences to these lapses, and unfortunately, moderation came late.
@moustafaayad @moustafaayad on x
1. As contributors to this report, it's important to bear in mind just how the spread of hate and terrorist content in the run-up to the Kenyan elections on August 9 stokes sectarian tensions. Here's how al-Shabaab and Islamic State sought to undermine the elections. Thread. http…
@kathleenklaus Kathleen Klaus on x
“'It is a deliberate choice to maximize labor and profit extraction, because they view the societies in the Global South primarily as markets, not as societies,' said Nanjala Nyabola, a Kenyan technology and social researcher.” https://www.washingtonpost.com/ ...
@g_bukeyeneza @g_bukeyeneza on x
Why dangerous content thrives on Facebook and TikTok in #Kenya | By @NehaWadekar https://www.washingtonpost.com/ ...
@moustafaayad @moustafaayad on x
This builds off the work of @ISDglobal and the deep dive into multilingual al-Shabaab and Islamic State networks across East Africa, and their proliferation on social media platforms, manipulating the information environment, spreading hate, polarization and extremism. https://tw…
@moustafaayad @moustafaayad on x
4. This is dynamic is prevalent anywhere there are jihadist groups are operating, and Kenya is no different. But the carte blanche they seemed to have on the platforms, in this context, was incredibly worrying.
@moustafaayad @moustafaayad on x
3. While violent posts were the most prevalent, second were posts that contrasted the Islamic State and al-Shabaab with a “corruption-riddled” democracy in bed with infidels. These posts sought to keep Muslims away from unIslamic polls, and choose jihad over the ballot box.
@mucheru @mucheru on x
Media, including social media, will continue to enjoy PRESS FREEDOM in Kenya. Not clear what legal framework NCIC plans to use to suspend Facebook. Govt is on record. We are NOT shutting down the Internet @NCIC_Kenya @citizentvkenya @Meta @MediaCouncilK @KenyaEditors @Reuters
@dellcam Dell Cameron on x
Facebook was informed it had approved pro-genocide ads in Kenya, so it took action by rushing out a press release describing all the ways it was cracking down on hate speech ahead of elections next month. Then it approved more pro-genocide ads. https://gizmodo.com/...
@katieharbath Katie Harbath on x
Kenya orders Meta's Facebook to tackle hate speech or face suspension https://www.reuters.com/...
@kofiemeritus Kofi Yeboah on x
The NCIC in conversation with the Communications Authority of Kenya (CAK), which regulates social media firms, recommended the suspension of Facebook ahead of the elections due to national security concerns. https://www.reuters.com/...
@mishichoudhary Mishi Choudhary on x
But it's not a western country, why should anyone care? It's only when Trump gets elected or something happens here that we should pay attention. https://twitter.com/...
@juliet_shen Juliet Shen on x
it's pretty clear to see when this happens as companies grow really big and inevitably adopt a strict top-down hierarchy to manage the behemoth of teams. But MAYBE that model doesn't work as well for teams whose missions typically align more with collective socialism & community
@blackamazon @blackamazon on x
Facebook regularly suppress racial / ethnic/ gender minorities Those marginalized folks notice and report YEARS before these “exposés” ... no one who goes back and asks those marginalized folks what's happened seems to catch any respect The. It's a SCOOP Lather rinse repeat https…
@noahjnelson Noah J. Nelson on x
If you have a problem like this the solution is straightforward: cut off serving ads in the affected area until you can assure this shit won't happen. https://twitter.com/...
@harry_jerry Harpreet Singh on x
Facebook once didn't approve my ad for @DealsForGeeks because it had too much text. https://twitter.com/...
@juliet_shen Juliet Shen on x
in my *personal* opinion, this is like how when regulators who don't fully grasp how tech works mandate certain things, it leads to more mistakes & lower quality decisions bc people who actually do the work race to make policy clarifications/workflow changes to support the optics…
@juliet_shen Juliet Shen on x
one of many reasons why I'm very into org structure and design and how that manifests in real world impacts. This is especially important in Trust & Safety, integrity, and anti abuse spaces.
@corinne_podger Corinne Podger on x
There's no way journalists can keep up with this level of misinformation, and unless they pay to boost, their content is unlikely to be seen at all on Facebook. https://twitter.com/...
@rweingarten Randi Weingarten on x
There is a clear need for oversight and consequences when Meta is encouraging genocide. There's no two ways about it. https://twitter.com/...
@simonplittle Simon Little on x
“Specifically, the groups said, Meta approved ads on Facebook in both Swahili and English that included calls to rape and behead Kenyan citizens along ethnic lines.” https://gizmodo.com/...
@christiancooper @christiancooper on x
I'm going to keep saying this: there is no economic value to Facebook above $0. Literally none. Imagine running a company and having a “pro-genocide” problem. https://gizmodo.com/...
@ianbetteridge @ianbetteridge on x
Tell me again about how Facebook is a powerful force for good, Nick Clegg? https://gizmodo.com/...
@marietjeschaake Marietje Schaake on x
The Kenyan regulator has given Facebook one week to remediate the approval of ads advocating for ethnic cleansing > Facebook's reply? ‘People and machines make mistakes’ > looks instead the company has not invested the resources or is simply incapable ↘️ https://gizmodo.com/...
@joanbarata Joan Barata on x
Platforms must be subjected to maximum scrutiny and transparency in contexts like this. But no matter how perfect human/automated/algorithmic moderation may be, there will always be errors at the intersection between “doing too much” and “not doing enough” https://gizmodo.com/...
@robertcaruso Robert Caruso on x
And this is after the puzzling yet proud and incessant messaging from Meta and IG proper folks that the inability to lay out simple photos chronologically on a grid is beyond their capability. https://twitter.com/...
@aweaverwrites Amanda Weaver on x
Facebook rejected one of my ads because the cover model had visible cleavage. https://twitter.com/...
@moonalice Roger McNamee on x
From today's edition of What Horrible Thing Did FB Do Today? https://twitter.com/...
@justinhendrix Justin Hendrix on x
“It was good to see positive trajectory on our engagement trends this quarter coming from products like Reels and our investments in AI,” said Mark Zuckerberg, Meta founder and CEO. “We're putting increased energy and focus around our key company priorities .... ” - Q2 earnings h…
@frankconniff Frank Conniff on x
And yet I was banned for life for making fun of Chuck Todd. True story. https://twitter.com/...
@safiyanoble Safiya Umoja Noble PhD on x
Facebook is an undeniable global hazard. What will it take to stop FB from promoting genocide and undermining elections in Kenya?
@robaeprice Rob Price on x
Last week, Facebook put out a press release touting how it has cracked down on harmful content in Kenya ahead of elections. Since then, it has approved pro-genocide ads, prompting a national watchdog to threaten to suspend it from the country. https://gizmodo.com/...
@femme_mal @femme_mal on x
How many genocides is this corporation going to have to aid and abet before people start pulling the plug on their Facebook/Messenger/Instagram/ Meta accounts? Will it take bloody civil war and genocide in the US before users clue in and leave? https://twitter.com/...
@blongstpaul Brandon Long on x
The Rawandan genocide was egged on by a radio station (Radio Télévision Libre des Mille Collines) @facebook is truly evil. https://twitter.com/...
@brooklynmarie Brooke Binkowski on x
They don't care. https://twitter.com/...
@hijakamran Hija Kamran on x
I'll keep saying this: Facebook paid $5B just to save Zuck in court - the money+more it could & can still spend to make its content moderation efficient in the larger world where most of its users are based. It deliberately ignores the harm it causes despite having all resources.…
@mehdirhasan Mehdi Hasan on x
Sheesh. https://twitter.com/...
@vermontgmg Garrett M. Graff on x
Serious question: How many genocides has Facebook been linked to around the world now? I've lost count, but anything more than zero seems ... not great. https://twitter.com/...

Chronicles