Twitter has been a hellscape these past few weeks for dissidents demanding accountability for the apparent murder of journalist Jamal Khashoggi — and a haven for pro-regime accounts harassing anyone who dares speak up. The question now is how the company can change that, and whether it even wants to.

The story of Saudi-sponsored Twitter propaganda starts with the Arab Spring, when governments around the world learned that the social media sites they had dismissed as little more than toys for teenagers could easily become instruments of rebellion. Autocrats set out to transform the tools of emancipation into tools of repression, hiring experts from across the ocean to teach them how. With enough money and enough manpower, dictators realized, they could replace narratives spread by dissidents with their own.

Saudi Arabia has a lot of money and a lot of manpower. The country’s population makes up a plurality of Twitter users in the Middle East, and the discourse on those channels is full of sectarian hate and “hashtag poisoning” to promote propaganda. Critics call Crown Prince Mohammed bin Salman’s strategist Saud al-Qahtani the minister of disinformation, or the minister of flies, which is how they refer to the gnat-like bot accounts who swarm the Twitter mentions of regime skeptics. Sometimes the flies spread misinformation. Sometimes they make threats.

It’s this automated army that activists identify when they ask Twitter to take action, and the company deserves accolades for removing a network of accounts this week, many of which were pushing pro-Saudi talking points about Khashoggi’s disappearance. But telling Twitter to tackle Saudi-sponsored harassment by banning bots misses an important part of the picture: The surest way stop an army is to take out its generals.

Much of the harassment directed against those the so-called Saudi enemies, it turns out, doesn’t come from anonymous hype-men with minimal followings. It comes from prominent people, with little blue check marks next to their Twitter names indicating that the site marks them as legitimate. Researchers have found that relatively few accounts are responsible for a relatively large portion of Twitter traffic — they tweet, and others retweet. Qahtani’s flies, it seems, need to know when and where to take wing.

Look at tweets about Khashoggi’s fiancee, Hatice Cengiz, for example. A verified account with more than 60,000 followers argued Cengiz is a man in disguise, surgically altered to pass as female. Smaller accounts pick up on the same thread and spread it.

Ghada Oueiss, an Al Jazeera journalist, has been called a “prostitute” by an account with 338,000 followers verified as Saudi Minister Abdullatif al-Shaikh. He forecasts her imminent demise in another missive. Other users, some of them likely bots, flood the replies to Oueiss’s tweets with similarly sexist rhetoric and edited images of her on the lap of Palestinian political philosopher Azmi Bishara.

“Very soon,” an account with 145,000 followers verified as pro-government blogger Munther al-Sheikh Mubarak said in response to another reporter’s tweets, posting a photo of the journalist alongside Khashoggi’s fiancee. The tweet’s defenders say the implication wasn’t that those targeted would be killed, only that they would be investigated and their lies exposed. This, in Saudi Arabia, is not comforting.

Verified accounts have also taken to disseminating lists online with the names of those they consider dangerous. Dutifully, the flies begin to buzz.

Some of these accounts are clerics close to the regime who systematically use their influence on Twitter to start trends supporting the monarchy. Others may want to demonstrate their loyalty in a society where patriotism has social and political currency. But they are more dangerous than any individual bot, because they have influence. And no one has as much influence as the maestro of online malevolence — Qahtani himself, who has createdblacklist of critics for Saudis to add names to.

Twitter is not blind to this abuse; when the subjects of harassment have reported tweets from verified accounts, some have disappeared from the site. But until the service seeks out violators on its own, and until it confronts the man they answer to, the infestation will remain.

It’s a big ask. Rooting out fake accounts is one thing. Rooting out people closely connected to the regime that has granted Twitter permission to operate in the country it rules — a regime whose members have reached into their purses to prop the company up — is another altogether. But a platform that made its name on its commitment to free speech has a responsibility not to abet tyrants and their cronies who skirt its rules to crush dissent.

“Block them and ignore them,” Khashoggi once messaged Oueiss, the Al Jazeera journalist, about the accounts harassing her. “Don’t let them make you sad.” Oueiss said she retweets her attackers to show the world what kind of people they are. Now it’s Twitter’s turn to show it is listening.