September 27, 2026
Political leaders across the West have said for a decade that governments must end online anonymity to shut down fake accounts known as bots, particularly those from Russia. “I want to see real names on the internet,” said German Chancellor Friedrich Merz in February. “In a democracy, citizens have the right to privacy,” said Spanish Prime Minister Pedro Sánchez at Davos in January 2025, and “not to anonymity or impunity,” proposing to tie every social media account to a European digital identity wallet. Greece’s digital governance minister proposed in April 2026 to link every profile to a verified identity because “The major problem behind anonymity is toxicity.” French President Emmanuel Macron declared that his priority was “to demand the disappearance of fake accounts” and “these famous bots, these troll farms.”
But now the advocates of government censorship of legal speech are demanding the legal power to wage information warfare through anonymous accounts and an explicit “sock puppet” strategy. Buried in the Australian government’s new censorship bill is a definition: “A sock puppet identity is a false or fictitious identity assumed by a person in the course of using, or carrying out any activity in relation to, an online service.” The legislation goes on not to warn of bad actors using sock puppets but rather to authorize the eSafety Commissioner, Julie Inman Grant, to create and use sock puppets to “disseminate” information.
Michael Nayna and Andrew Lowenthal first reported on the sock puppet provisions last week.
Defenders of the legislation say the sock puppet power is limited to research and regulatory testing. “This is just not true,” wrote Cameron Wilson, an Australian Broadcasting Corporation (ABC) reporter on X, referring to the sock puppet language. He said “the legislation limits the sock puppet powers to ‘performing the Commissioner’s functions or exercising the Commissioner’s powers’ and may only ‘engage with users only to the extent necessary to prevent closure of the account.’” The section “legitimises and protects certain uses of accounts run with false identity information in order to test features and protections that digital platforms claim to have put in place,” said Tama Leaver, a professor at Curtin University.
But the legislation explicitly says “sock puppet,” a term for spreading deception, and it does not limit the government’s use of anonymous accounts. The legislation says “subsection 205J(1) authorises (without limitation) the use of sock puppet identities by the Commissioner to generate material using an online service (including by means of artificial intelligence) to the extent that the production, dissemination or possession of that material is not an offence against a law of the Commonwealth, a State or a Territory.” A proposed AI-created Community Note for Wilson’s post points to the “without limitation” language.
Not only that, but the legislation explicitly says the government will not have to follow the rules against the use of anonymous bot accounts on social media platforms. The law gives the government powers to operate “despite anything to the contrary in” any other law and in “any agreement, contract, deed, policy or other document.” The government is thus proposing to undermine X’s authenticity policy, which bars accounts “that are not legitimate, genuine and transparent as to their source, identity, and popularity.” And Australia’s industry code on disinformation commits platforms to act against “the use of fake accounts or automated bots that are designed to propagate Disinformation.”
The Australian government’s demand to essentially spread government propaganda through anonymous bot accounts is the exact opposite of what the same government demanded in a failed censorship bill two years ago. Back then, the proposed misinformation bill defined “inauthentic behaviour” it wanted platforms to punish as dissemination that “uses an automated system (for example, an algorithm or bot network) in a way that is reasonably likely to mislead an end user about a matter such as the identity, purpose or origin of the person disseminating the content.”
Why is that? Why is the Australian government demanding the right to, at a minimum, spread government propaganda anonymously and, more maximally, wage offensive information warfare?
Bot Projection
In 2016, the people who went on to build the censorship industrial complex raised the alarm that Russian bots had swung the American election. “Experts are telling us that Russian state actors broke into the DNC, stole these e-mails,” said Robby Mook, Hillary Clinton’s campaign manager, on CNN on July 24, 2016, “for the purpose of actually helping Donald Trump.”
Nobody raised the alarm about anonymous accounts more than Renée DiResta, who would later lead censorship advocacy at the Stanford Internet Observatory. In 2016, she explained, “On the Internet, nobody knows you’re a dog. Or a Kremlin bot. Or a paid commenter.” In May 2017, she wrote that “Disinformation-campaign material is spread via mass coordinated action, supplemented by bot networks and sockpuppets (‘fake people’).”
The US Intelligence Community (IC) concluded in a 2017 assessment that Russia’s campaign used “paid social media users or ‘trolls’” from the Internet Research Agency.
“Facebook and Twitter remain littered with pro-Russian, Western-looking accounts and supporting automated bots designed to undermine the credibility of the U.S. government,” testified Clint Watts, a former FBI agent, to the Senate Intelligence Committee on March 30, 2017. That August, the Alliance for Securing Democracy launched a dashboard called Hamilton 68 and said its analysts had “linked 600 Twitter accounts to Russian influence activities online.” A bot, the Atlantic Council’s Ben Nimmo wrote in 2017, is “simply an account run by a piece of software, analogous to an airplane being flown on autopilot,” and bots “pose as real people.”
DiResta worked at a partisan political consulting firm, New Knowledge, that used anonymous accounts to wage bot warfare against a Republican. An internal New Knowledge document on the 2017 Alabama Senate race said that “We orchestrated an elaborate ‘false flag’ operation that planted the idea that the Moore campaign was amplified on social media by a Russian botnet.”
And yet the Senate Intelligence Committee released New Knowledge’s report on the Internet Research Agency in December 2018 with DiResta as first author under the pretense that it was a nonpartisan report. DiResta said she was uninvolved, telling the Times, “I know there were people who believed the Democrats needed to fight fire with fire.”
But then, in 2018, DiResta disclosed that she herself ran anonymous accounts. At the Aspen Ideas Festival that June, on a panel titled “Anti-Social Media and The Menace of Disinformation,” she described how she studied Facebook’s recommendation engine.
“I have a bunch of accounts that pay attention to anti-vax content,” said DiResta, “and they were getting served pizzagate groups long before Pizzagate was a massive public story.”
This provoked the moderator, Nicholas Thompson, then the editor of WIRED and now the chief executive of The Atlantic, to ask “How many bot accounts do you run?”
“No comment,” DiResta responded
“I would like a complete tally of Renee DiResta’s sock puppet accounts by the end of this panel,” teased Thompson.
DiResta wasn’t the only person on the panel who admitted to doing it. Kashmir Hill of Gizmodo said, “I create sock puppets on Twitter.”
Thompson then asked “if anybody has questions either from themselves or their sock puppets.”
As such, one possibility is that DiResta and others on the left operated fake accounts to swarm their enemies during the years they accused their enemies of doing the same, a strategy that began with the Clinton campaign’s 2016 decision to paint Trump as a Russian asset.
The best way to understand people like DiResta is as partisan professionals linked to the intelligence world who worked through informal, nongovernmental networks, from the dismissal of the Clintons’ influence peddling to the Russia collusion story to Covid, election, and Hunter Biden censorship.
Support for this comes from Hamilton 68, which investigative reporter Matt Taibbi exposed in a Twitter Files thread. Inside Twitter, executives viewed Hamilton 68 as untrustworthy. “I think we need to just call this out on the bullshit it is,” wrote Yoel Roth, Twitter’s head of trust and safety, in internal messages, after his staff found the dashboard’s accounts were “neither strongly Russian nor strongly bots.” The accounts were 86% English-language and mostly real people in the United States, Canada, and Britain.
Hamilton 68’s creator worked for the FBI and DiResta worked for the CIA. She describes her own start as “an undergraduate internship with the CIA” and in 2025 she said she “ interned for the CIA” from “1999 to 2024.”
As research manager of the Stanford Internet Observatory (SIO), DiResta led the Election Integrity Partnership (EIP) in 2020 and the Virality Project in 2021. The EIP was “Created in the summer of 2020 ‘at the request’ of the Cybersecurity and Infrastructure Security Agency (CISA),” found the House Judiciary Committee in 2023. “What the federal government could not do directly, it effectively outsourced to the newly emerging censorship-industrial complex,” the report noted.
The censorship industrial complex is comprised of left-wing NGOs like SIO that demanded censorship, academic researchers, the National Science Foundation (NSF), and others.
Since 2011, the National Science Foundation has spent at least $23.9 million on about 50 grants to study, detect, and in some cases operate the automated and fake accounts used to shape opinion online. One of the earliest was a 2014 grant to the University of Washington’s Philip Howard for $219,000 for “Computational Propaganda and the Production/Detection of Bots.” It paid for fieldwork with the political consultants hired to build bots, and for new tools to catch them. In 2018, NYU’s Joshua Tucker received$692,000 in 2018 to develop methods for tracking “domestic and foreign” bots in political discussion.
In 2020, NSF funded a workshop on “Inauthentic Behavior in Online and Digital Systems” to scope a new track of its Convergence Accelerator program. That track, “Track F,” then funded several projects. They include $5 million and $750,000 for SUNY Buffalo’s deception and “Disinformation Range” training, an MIT program “to combat inauthenticity”, and University of Wisconsin tools for identifying networks spreading election and vaccine misinformation.
Boston University and Binghamton University received in 2021 about $500,000 to build machine-learning models of “troll” accounts, trained on accounts that Twitter and Reddit had attributed to foreign governments. The same team, joined by the University of Vermont, received in 2024 another $600,000 to use large language models to find “inauthentic accounts.”
Smaller NSF grants funded defenses against “sockpuppet” friend requests, covert social bots, fake accounts, and paid “crowdturfing” campaigns. Even when these grants are framed as defensive against bots, they provide the knowledge for using them offensively. UC Davis received $300,000 for work tested through “sock puppet-based experiments” on recommendation algorithms. Duke’s Christopher Bail received funding for a Twitter field experiment that exposed users to opposing views; the grant abstract doesn’t mention bots, but the published study used bot accounts. Bail later received funding to train large language models to act as simulated social media users. Cornell and Princeton received $800,000 for a platform where thousands of humans interact with AI “bots”.
Under one of its contracts, researchers at Utah State University and IBM created “a total of 17 Twitter accounts and our system automatically sent retweeting requests to 3,761 strangers on Twitter,” nine of them under the name “Public Safety News.” The study said that 8.4% of the strangers the bots messaged about bird flu retweeted them.
Given this history, it is reasonable to assume that the Australian government wants the power to operate networks of AI-fueled bots in order to wage information warfare against its political enemies. The Australian government has repeatedly made clear that it wants more censorship powers because it fears the rise of an anti-mass migration political party called One Nation. It now appears that running an offensive information operation is part of that effort.
DARVO
Now that they have been caught, the censorship advocates say there was never any censorship. Anne Applebaum in The Atlantic wrote on September 2, 2026, the censorship industrial complex “belongs to that mystical category of things that disappear when you examine them too closely.” DiResta and the Stanford researchers “had merely noted their existence” of 400,000 tweets, “after the election, not in some kind of smoke-filled room but in an academic report.” Eileen Guo, Gisela Perez de Acha, and Martin Sona wrote in the MIT Technology Review, “Much of what has been characterized as censorship is really part of an effort to make the internet safer.”
The individuals deny their own involvement in censorship, which is well established. DiResta admits that “undergraduate students and researchers flagged ~4800 URLs to platforms that ignored ~65% of them and took down ~10%.” DiResta’s counts 480 URLs taken down at the request of a project whose creation the House Judiciary Committee documented as “at the request” of a federal agency, and whose own final report said that 35% of the links it sent to platforms “were either labeled, removed, or soft blocked.”
The heads of the social media companies said Democrats pressured them to censor. Senior Biden administration officials, “including the White House,” “repeatedly pressured our team for months to censor certain COVID-19 content, including humor and satire,” Mark Zuckerberg wrote to Congress in August 2024.
Applebaum has worked with a British military and intelligence government program. The Foreign and Commonwealth Office funded the Institute for Statecraft’s Integrity Initiative with £296,500 in 2017-18 and “a further £1,961,000” in 2018-19, Sir Alan Duncan told Parliament in December 2018, for a program “to defend democracy against disinformation.” The institute’s leader, Chris Donnelly, had headed the British Army’s Soviet Studies Research Centre at Sandhurst, and its 2018-19 application to the Foreign Office named “the LSE team,” which academics who analyzed the leaked documents said “appears to include Anne Applebaum and Peter Pomerantsev.”
And so, when exposed, censorship advocates engage in DARVO, which stands for “deny, attack, reverse victim and offender.” The people who silenced Americans are claiming that the people who exposed their efforts are trying to silence them.
NSF’s Convergence Accelerator “Track F” funded 12 teams in September 2021, including one at the Massachusetts Institute of Technology (MIT) titled “Adapting and Scaling Existing Educational Programs to Combat Inauthenticity and Instill Trust in Information.” The NSF awarded Adam Berinsky and David Rand $881,188 in 2021 for “Promoting Accurate Information on Social Media,” a grant the agency terminated in April 2025. An MIT team was among the 12 Track F awardees the NSF announced in September 2021.
MIT is also the source of the most quoted pieces of misinformation used by censorship advocates in the form of a 2018 Science paper that purportedly found that “Falsehood diffused significantly farther, faster, deeper, and more broadly than the truth.” But the roughly 126,000 tweets that MIT researchers analyzed to inform the study’s findings are equivalent to the number of tweets published in a mere 21 seconds today. In other words, they generalized from 21 seconds of tweets to the whole of the Internet to make their sweeping claim.
Other government officials who engaged in censorship and information warfare for partisan purposes also denied what they did when caught. GCHQ’s Joint Threat Research Intelligence Group listed among its tactics in 2014 “’false flag operations’ (posting material to the internet and falsely attributing it to someone else).” When caught, governments deny what they’ve done. The British government said in January 2021 that the Army’s 77th Brigade “do not, and have never, conducted any kind of action against British citizens,” after the brigade had spent 2020 monitoring the Covid posts of British citizens. And Canada’s military ran an information operation on Canadians that its own acting chief of the defence staff said was conducted “without explicit CDS/DM direction or authority to do so.”
The Australian government’s new information warfare and censorship bill suggests Australia could be a proving ground for efforts at information control by a network of liberal and left-wing officials who move between intelligence agencies, governments, and think tanks. Inman Grant told TIME in October 2025 that her other job offer out of college “was with the CIA.” In April 2024 she demanded that X delete a video of a Sydney stabbing not only in Australia but worldwide. In September 2025 she gave the keynote at a closed Stanford meeting where officials from the European Union, Britain, Brazil, and Australia met American censorship advocates to coordinate.
The US Embassy in Canberra called the Australian bill “extraterritorial censorship of protected speech” on September 22. But it did not mention the bill’s provision for Inman Grant to secretly “disseminate” information to potentially partisan or ideological ends.
The lesson is that everything censorship advocates say about their enemies is a tell about what they are doing themselves. They warned of Russian bots while their own firm planted a Russian botnet story in Alabama, and their own researcher declined, on stage, to count her fake accounts. They demanded that platforms ban “fake accounts or automated bots” and now write a statute that bars platforms from closing the government’s. And they call for the end of anonymity for citizens and immunity while they seek to use anonymity for themselves.
