On October 10th, as members of a satirical-turned-serious youth movement gathered in New Delhi to demand the resignation of India’s Chief Election Commissioner, the Ministry of Electronics and Information Technology sent Meta an advisory. It did not ask the company to delete anything. It asked Meta to keep its recommendation systems from pushing a “limited set of narratives.” Three months earlier, in Brussels, the European Commission had told the same company that the way its apps recommend content may break European law. Two governments, two continents, one target: not a post, not an account, but the machinery that decides what we see. Their demands look alike. They are not, and the difference is worth getting right.
The word at the center of both fights has an old pedigree. In the ninth century, the Persian mathematician Muhammad ibn Musa al-Khwarizmi wrote a treatise on reckoning with Hindu numerals; Latin translators rendered his name Algoritmi, and the word that grew from it came to mean a procedure, a sequence of steps from problem to answer. Long division is an algorithm. So is a recipe. For most of its history, an algorithm was something you followed.
Somewhere in the past decade, the grammar flipped, and the algorithm began following us. In the spring of 2016, when Instagram announced that it would stop showing posts in chronological order, more than three hundred thousand users signed a petition against the change. The protest failed, and the phrase that emerged from it stuck. People stopped saying “an algorithm” and started saying “the algorithm,” with the definite article we reserve for the weather or the market.
The Indian case begins with a joke. The Cockroach Janta Party was launched online in May by the political strategist Abhijeet Dipke, as satire aimed at the political establishment. Within months it had become a channel for the political and economic frustrations of young Indians, and its protests had spread from Delhi to other cities. By autumn, its demand was specific: the removal of Chief Election Commissioner Gyanesh Kumar over the revision of the country’s voter rolls.
The ministry’s advisory, which became public through media reports on October 18th, asks Meta to offer a diversity of relevant and credible information, and to act immediately and report back. Ministry sources told the Press Trust of India that agencies had flagged Meta’s algorithms as appearing to favor one kind of content. The government’s stated concern is public order. Officials point to old videos circulating as if they were current, with unverified claims about police actions feeding public anxiety, and they want context attached to images whose date, location, or authenticity is unclear. That is a reasonable worry; misattributed footage has inflamed real crises before.
But the advisory also asks Meta to prioritize complaints from government agencies, and it arrived alongside more than two hundred takedown orders issued by Delhi Police during the protests. Digital-rights advocates have called for the advisory to be withdrawn and for Meta to publicly disclose every removal and downranking carried out under it. Neither the ministry nor Meta has published such a list.
The European case is drier, and slower. On July 10th, after an investigation opened in May 2024, the Commission preliminarily found that Meta had violated the Digital Services Act through the addictive design of Instagram and Facebook. It concluded that the company had not properly weighed the effects of features such as personalized recommendations, autoplay, and infinite scroll on users’ wellbeing, including minors’. It found that Meta had disregarded what it knew about how much time minors spend on its apps at night. The Commission wants autoplay and infinite scroll off by default and recommendation systems less focused on maximizing engagement. Meta rejected the findings, saying that they don’t reflect the steps it has taken to protect teens. The case is not over, and a fine, if one comes, could reach six per cent of the company’s global annual turnover.
Put the two actions side by side and they share a premise that would have sounded strange a decade ago: that a recommendation engine is a kind of editor, and that governments are entitled to opinions about its judgment. The difference lies in which part of the editor they want to change.
Brussels is objecting to an appetite: the system’s hunger for attention, regardless of what holds it. A rule that autoplay must be off by default applies equally to cat videos and to protest footage. It can be written down, tested, and argued over in court, as Meta is now doing.
New Delhi is objecting to a choice of story. A demand that an algorithm stop favoring a “limited” narrative requires someone to decide which narrative is limited, and how much of it is too much. During a protest against an election official, the party making that judgment is the government whose official is under fire. The concern about doctored or misdated footage is legitimate. The remedy, as drafted, gives the state a hand on a dial whose settings no one outside can see.
That opacity is the deeper problem, and it predates both governments’ interventions. When the public cannot see why a post rises or sinks, anyone with power can claim to know, and the platform can deny anything. Brussels has tried to address this with mandatory risk assessments and audits. India’s advisory asks Meta to report back to the ministry, not to the public.
Al-Khwarizmi’s procedures had a virtue we have since misplaced: they were public. Anyone could check the steps. The most useful demand to make of the feed, from protesters, parents, or ministers, may be the oldest one a mathematics teacher ever made. Show your work