From Turkey to India, Twitter offers censorship on demand

Ellery Roberts Biddle


When governments come knocking, Twitter is glad to censor. Under Elon Musk, the bird company is honoring most government demands to take down tweets or hand over users’ data, according to data from Lumen and a new report from our friends at Rest of World.

It wasn’t like this before. Publicly available records suggest that the company is fully complying with 80% of government requests, in contrast to the pre-Musk era, when that number hovered around 50%. This was thanks in part to policy staffers who worked hard to figure out when government demands were really legitimate and when they were overblown. But Musk fired most of the people doing this work shortly after taking the helm. And now people around the world are feeling the consequences.

More on this below.

With the “fake news” bill looming, Brazil blocked Telegram over neo-Nazi channels.

Brazil’s Congress is poised to vote on a controversial, Bolsonaro-era “anti-fake news” bill that would require big tech platforms to proactively remove illegal content, curb mass messaging by politicians and make other changes that would give the government more leverage when dealing with foreign tech powers. Silicon Valley wants none of it — Google even used its quasi-monopolistic online presence to push its agenda in Brazil — and some free expression advocates are concerned too.

As if to offer a case in point, major internet providers in Brazil blocked Telegram on April 26, over the Dubai-based company’s refusal to hand over information about neo-Nazi activity on the platform. Police had requested data about two groups they suspect used Telegram to encourage a series of violent attacks at schools in Brazil in recent months. Telegram says the data can’t be recovered. The block on Telegram was lifted on May 2, but the company is still racking up fines to the order of $200,000 per day. For Telegram’s part, the company says it has never complied with a single data request from a government or anyone else. If the fake news bill passes, this may have to change, or Telegram may need to say, “bye bye, Brazil.”

African content moderators are uniting. A coalition of workers who clean up troublesome content for some of the world’s largest internet platforms — Meta, ByteDance (owner of TikTok) and OpenAI — voted to form the African Content Moderators Union this week. This is the latest development coming out of legal battles in Kenyan courts over the rights of content moderation workers who are typically hired by third-party companies — Sama and Majorel are two of the biggest players in Nairobi — that offer low pay and next to no benefits. Happy May Day, folks.


So now we know: Twitter is taking governments at their word and removing most of the tweets they say are illegal — and maybe some that just rub government officials the wrong way. Meanwhile, emerging evidence shows that the company is less interested than ever in proactively removing content that violates its own policies, not to mention content that violates local laws. I’m thinking here about all the violent, hateful and otherwise nasty stuff that third-party content moderators have to deal with. At least the ones in Kenya might have union representation soon.

I asked Turkish internet law scholar Yaman Akdeniz about it this week — Turkey has made more censorship demands of Twitter than almost any other country on earth. Akdeniz noted that in the past, it was clear that the company “ignored” most requests from Turkey. “I am not sure if this will be the case with the Musk administration,” he said. The data from Lumen certainly suggests that it won’t.

He described how Turkish authorities restricted access to Twitter following the earthquake earlier this year. The response from Twitter was swift. “There was an immediate meeting,” he said, and the ban was lifted within hours. 

“I can only speculate what was promised in that meeting,” he said. “More tweets will be withheld and more accounts will be suspended, that is for sure.” And none of this bodes well for national elections, which are coming up on May 14. “Twitter can easily become the long arm of the law enforcement agencies in Turkey if AKP wins,” Akdeniz warned.

India was also at the top of the list of governments asking the company to take down tweets. In March, we wrote about Twitter’s willingness to censor tweets about the police search for a Sikh secessionist preacher in Punjab. The episode made it look as if the company was glad to do whatever the Indian state or federal authorities asked, including suspending the account of a member of the state assembly. 

If government officials can simply lean on Twitter to silence not only their critics in the public sphere but also their political opponents, the consequences for public discourse — and democracy — will be pretty severe. 

Censor when governments ask, but let the rest flow as it will. What could go wrong?


  • Israeli authorities have added facial recognition technology to their arsenal of surveillance tools used to target Palestinians in occupied territories. Amnesty Tech has a new deep dive on the Red Wolf tool.
  • It’s common in the West to think of China’s internet censorship regime as a monolith, but with a tech industry as big as China’s, controlling online information is complicated and not always consistent. Citizen Lab has a new study on search engine censorship in China that digs into the details — it is worth a read.

From biometrics to surveillance — when people in power abuse technology, the rest of us suffer

More Coda Newsletters