X Suspends 13,000 Foreign Influence Accounts Before US Midterms

A crowded week for X

In the third week of October, X made news in several areas at once. It removed thousands of accounts it linked to foreign governments. It changed how it sells subscriptions, set new rules for advertisers and had outages in some regions. Meanwhile, its owner kept spending heavily on US elections. Taken one at a time, each item is ordinary platform news. Taken together, they show a company trying to look like a neutral, reliable "town square" while being pulled more tightly into Elon Musk's other businesses and his politics.

This article argues that the account takedown is the week's most important story. It is not the biggest operational change. It matters because it shows the tension that now defines X most clearly.

The takedown: 13,000 accounts, three countries

X said it had suspended more than 13,000 accounts tied to three suspected foreign influence operations linked to Iran, Russia and China. According to the company, all three were trying to shape political debate in the United States1. X announced the action on Wednesday, October 14, together with its plans for the November 3 midterm elections119. The company said the networks ran fake personas that posted and boosted political content aimed at candidates and public figures "across the political spectrum"119.

X said it found the accounts through proactive sweeps and investigations under its rules on spam and coordinated inauthentic behavior1. Its government affairs team presented the move as a free-speech commitment, arguing that X's job is to keep the conversation authentic, open and safe1.

The size of the takedown stands out against the platform's own history. In 2021, when it was still called Twitter, the company announced the suspension of 238 accounts allegedly run from Iran115. A single announcement 55 times larger suggests that influence campaigns have grown, that detection has improved, or both.

The timing also matters. X's announcement came about a week after OpenAI said it had shut down two influence operations from Russia and Iran. Those operations had used ChatGPT to invent fake journalist personas and to run a front "think tank" spreading anti-Ukraine propaganda1. IranWire, citing OpenAI's October 8 report, said one network used AI and fake personas to write or repost nearly 100 articles under seven pseudonyms across 12 digital outlets, posing as Western journalists19. Coverage of both announcements points the same way. Influence operations now run across several platforms, generative AI makes fake personas cheap to produce, and platforms are under pressure to show results before voting begins.

Election rules, Community Notes style

X also explained how it will police election content. It bans misleading claims about how, when or where to vote, efforts to suppress turnout, and voter intimidation, including calls for violence near polling places1. Users can report posts through a "civic integrity" option, which in practice asks for a Community Note to add context. In urgent cases, X may label posts that break its election rules1.

The design says a lot about X's approach. Relying on crowd-sourced context, with labels kept for serious cases, matches the platform's long-standing dislike of heavy-handed moderation. That system has its own baggage. Musk complained in early 2025 that Community Notes was "increasingly being gamed by governments & legacy media" and said he would "fix" it. He made those comments after notes had contradicted some of his own claims14.

The owner problem

The election safeguards land awkwardly because of who owns the platform. The Guardian, citing figures first reported by the New York Times, reported that Musk has become the biggest political donor of the 2026 midterm cycle, putting about $366 million into political groups and campaigns that are mostly on the right10. More than $276 million of that has come since July, which is more than he gave to support Donald Trump in 202410. The same report says Musk has long used X to attack political opponents and push election misinformation. It notes that over a recent weekend he reposted an America PAC video about a Democratic candidate's past comments on gender10.

The criticism extends beyond the US. Britannica's news feed notes a Guardian report in which a French minister accused Musk of running "online harassment campaigns" against European leaders12. Musk has also used X during the Starlink dispute over Ukraine. CleanTechnica reported that he used the platform to blame the US president for a decision limiting service, and that the White House later reportedly denied any responsibility9.

Here the coverage splits in a real way. X's company accounts describe a neutral referee policing fake accounts "across the political spectrum"1. The Guardian describes the platform's owner as one of the most aggressive partisan players in the same election10. Both can be true. Taking down state-backed fake accounts and an owner spending to win the Senate are separate questions. Still, the credibility of X's integrity work now depends on users believing that the company's enforcement is separate from its owner's politics. This week's coverage gives readers reasons on both sides.

Monetization: one bundle, stricter ads

On the business side, X changed both how it charges users and what it will accept from advertisers.

Teslarati reported that SpaceX Super Intelligence has launched "The X Subscription," a plan that combines Grok, Grok Bot, the AI coding tool Cursor and X Premium3. Voice of Emirates described the offer as four tiers aimed at individuals, developers, professionals and heavy AI users4. These accounts don't actually conflict. One umbrella product with tiered pricing fits both. The more important point is strategic: the social network is no longer sold on its own but as the front door to an AI bundle.

On Friday, Musk posted that all advertising on X will now be held to "a reasonable esthetic standard," and that advertisers whose ads are "ugly, shrill, strident" or overstate product claims will be barred5. The move follows his June 2025 ban on hashtags in ads5. VINnews noted that Musk gave no enforcement criteria, no start date and no triggering incident, which leaves X with wide discretion5. A platform still trying to win back ad spending has decided to turn some advertisers away. That is either a confident bet that better-looking ads earn more over time, or one more case of policy announced in a single post and filled in later. My reading is that it's mostly the second, given how little detail came with it.

Whose company is X now?

Even basic descriptions of X's ownership differ, which shows how fast its corporate structure has changed. Britannica says Musk moved X into xAI in March 2025 in an all-stock deal valuing X at $33 billion, and that SpaceX acquired xAI in February 2026 before a planned IPO, making X an indirect subsidiary of a publicly traded group12. Wikipedia lists the owner as "SpaceXAI," SpaceX's AI subsidiary11. Teslarati reports that Musk announced on October 4 that SpaceXAI would become SpaceXSI, and that the new subscription page already uses that name3.

The gaps come down to naming and timing more than substance. All of these accounts agree that X now sits inside SpaceX's AI business. That matters for how to read the platform. An analysis piece from ABAB News argues that resources are moving from content moderation toward supplying data and real-time training for Musk's AI efforts, turning the social network into a pipeline for xAI and SpaceX2. That is interpretation, not reporting, but the subscription bundle supports it.

Is X recovering?

Opinions on X's health are as divided as ever. ABAB News summarized John Gruber's argument on Daring Fireball that X has regained its footing. Gruber says discussion there, especially about AI, is livelier than on Threads, Bluesky or Mastodon, and that many accounts he follows still post only on X2. The same piece says most top advertisers have returned, even if ad revenue hasn't reached pre-takeover levels2. That matches Britannica's earlier note that advertisers came back without restoring their old budgets12. Musk also used Sundar Pichai's announcement of a Gemini model on X to argue that CEOs posting directly beats press releases8. Musk himself shared the Tesla Roadster demonstration video on X6.

Reliability is still a weak point. Users in India reported login failures and feeds that wouldn't load on October 14, with complaints peaking around 10:30 a.m. and 12:30 p.m. local time13. Separate monitoring data recorded a 37-minute incident that morning, along with US reports of black screens and failed posts.

The bottom line

The week's coverage describes a platform that is commercially more coherent than it was a year ago and politically harder to trust as neutral. X is showing it can find and remove state-backed fake accounts at scale, and it is building a clear subscription business around AI. Its owner, meanwhile, is the cycle's largest donor and one of its loudest partisan voices on the same platform. Until X separates its referee role from its owner's campaigning, each integrity announcement, including a real one like this week's takedown, will be read through Musk's politics.