I am texting with Mo, who calls himself an activist but looks a lot like a citizen journalist. He’s on the ground in one of the many Iranian cities experiencing waves of protests in reaction to the government’s May 6 announcement that food prices will rise between five and 10-fold—a spike that will have massive repercussions in a country already facing an economic crisis, inflation, and poverty.
Mo (which isn’t his real name) tells me he’s sitting in the home of relatives outside the city center to use their broadband home network. Usually, like the majority of Iran’s population, he connects to the internet through mobile. When the mobile internet is shuttered, as increasingly happens during protests, it is a near total internet shutdown for the typical user in Iran. Over the past few days, in the city he’s in, mobile internet connections have at times been completely shuttered. At other times, they are shuttered except for access to national applications of internet services, or thoroughly throttled so as to make them virtually unusable. “I can’t open WhatsApp without seeing ‘connecting,’ ” he tells me.
In December 2017 and January 2018, protests against corruption and a sinking economy overtook Iran. Since then, whenever protests erupt, armed security forces head to the streets, while authorities work with internet service providers to shutter or throttle access to the international internet. During the November 2019 protests against fuel price increases, the government blocked both the international and national internets. After the first 24 to 48 hours of the shutdown the authorities restored the national internet, while international connections remained shuttered for six to 10 days, depending on how quickly officials stifled local protests. (The national internet, known as the National Information Network, hosts national Internet infrastructure such as national banking and e-governance systems; it also includes communication and information services that adhere to government censorship and surveillance practices.)
Today, as Mo’s experience shows, the shutdowns have remained mainly on mobile networks in the localized areas of protests, even after renewed protests following the tragic collapse of Abadan’s Metropol building on May 23. Some evidence of nationwide disruptions have emerged, but nothing has erupted to the scale of November 2019. At least not yet. So Mo is able to use his relatives’ valuable broadband connection. But he and other activists are encountering censorship from an unlikely source: Instagram.
At one point, Telegram was the main communication tool during protests. But in May 2018, the app was censored by Iranian authorities. Now, Instagram is Iran’s most popular and only uncensored foreign social media platform. (It’s the second most used app after WhatsApp.) And in recent weeks, it’s begun taking down footage of protests and related content, apparently because of a policy change on not administering exceptions in reaction to the backlash against Meta’s content moderation policies at the start of Russia’s invasion of Ukraine.
On May 12, reports started to surface that users posting about the protests in Persian were experiencing mass takedowns. The affected Instagram accounts included one of the biggest protest documentation networks, the 1500Tasvir collective, and even in one case the diaspora Persian language media outlet Iran International.
All of the content removed appeared to have one thing in common: either a caption or audio included the common dissident protest slogan “Death to the Dictator,” reframed to include Iran’s du jour cadre of dictators including: “Death to Khamenei” (the current supreme leader); “Death to Raisi” (the current president) and “Death to the Revolutionary Guards/Basij” (the paramilitary forces responsible for violent repression of protest and dissent).
To Western observers, it might seem obvious that “Death to” a person would violate content guidelines against calling for violence. But in the Iranian context, “death to the dictator” has long been a symbolic slogan of dissent against Iran’s theocratic authoritarian system, rather than a call for actual death. At one point, Meta—Instagram’s parent company—understood this. During the July 2021 protests, after much reporting and discussion, Facebook created a “death to Khamenei” temporary exception to content moderation guidelines. Now, almost a year later, the same problem has emerged, only the scale of the protests have expanded and Meta is no longer abiding by that exception.
It appears that this is a result of criticism for creating special exemptions around the invasion of Ukraine. In March, Reuters leaked that Meta had created a number of exceptions that would allow Ukrainians to post expressions of violence toward Russian soldiers or invaders. It also created a special exception to allow praise for the the Azov Battalion, a neo-Nazi military group designated that at one point was considered for inclusion on the U.S.’s Foreign Terrorist List and subsequently Meta’s Dangerous Individuals and Organizations list.* The exception was given to protect speech that praised the Azov Battalion in the context of its efforts to fight the Russian forces.
These exceptions received a lot of backlash, especially because the public only became aware of them because of leaks, rather than announcements about transparent policies. Many critics focused on the fact that Meta and other platforms were giving far more care to their approaches to Ukraine than for similar situations outside of Europe, including in contexts such as Israel and Palestine or Syria.
But Meta appears to have taken the wrong lesson from this backlash. The problem was never that it should not have made exceptions for Ukraine. Yes, the wording and communications around the policies likely could have been subject to more thought and care. However, the solution to the Ukraine backlash should not be to end the ability to give exceptions.
When I asked Meta for comment about this, a spokesperson told me “We support free political expression, including criticism of leaders and governments, but not calls for violence or death.” They offered no information on the use of exceptions despite their well-documented precedence.
Meta’s deficiencies in content moderation in non-Western languages on their platforms range wide. Beyond the protest-related takedowns, Meta has misapplied policies as well as made numerous unexplained “mistakes.” It was recently reported that Meta’s Persian content moderators have been offered bribes to deplatform high-profile activist accounts like Masih Alinejad. While the unnamed moderator who recounted the story said they didn’t accept the bribe, the implications of such overreach by the Iranian government have stirred deep unease among Persian language users. The company has allegedly pledged to do an investigation, though it’s unclear if the findings will ever be made public.
Iranians have also been frustrated by the misapplications of Meta’s “Dangerous Individuals and Organizations” policy. The purpose of the policy is to avoid glorifying those included on the list. But the policy has also been used to remove content criticizing people and groups on the list. I’ve worked with dissident exiled rappers who have had their album art work aimed at criticizing the Revolutionary Guards removed repeatedly. I’ve seen accredited media organizations posting Persian news stories related to the Revolutionary Guards get so many mistaken takedowns of their news stories, their accounts were on the brink of being completely suspended. Sometimes the problem is related to using “media matching banks” for automated removals of images related to groups on the Dangerous Individuals and Organizations list, on which Iran’s Revolutionary Guards are prominently included. Other incidents have demonstrated content moderation systems incapable of understanding the context for satire or news. The massive numbers of mistakes demonstrate a need for an overhaul of these Persian moderation systems to meet the responsibility for expression they hold in Iran. Distrust of these misapplications are rampant. Perhaps most upsetting of all, for days Instagram blocked the use of a hashtag intended to pay tribute to the victims of a Ukrainian airliner shot down by the Revolutionary Guards over Tehran in January 2020. Meta explained the issue as “glitch.”
Mo, has been doing his best to ensure the voice of Iranians opposed to the Islamic Republic of Iran comes to the fore, but his frustrations at seeing protest images removed is palpable. “It seems we are always going to be caught in this horrible net,” Mo texted me in the days following the 1500Tasvir mass takedowns. “We’re between the brutality of this regime and then powerful companies like Meta who just don’t seem to care about us when we are crying for help.”
While misapplications of policies and a refusal of contextual analysis prevails, users are currently self-censoring to get by on a platform that holds monopoly over the majority of public online expression in Iran. One of the recent protest posts by the 1500Tasvir group is captioned as follows: “Please listen to the chant, if we write it out we will get censored.”
Correction, June 6, 2022: This article originally misstated that the Azov Battalion is part of the U.S. Foreign Terrorist Organizations list. It was considered for inclusion but not officially listed.
Future Tense is a partnership of Slate, New America, and Arizona State University that examines emerging technologies, public policy, and society.