Disinformation Wars: The Extended Edition

Want to see several people argue over *extremely granular* detail? This is your lucky day!

It has been quite some time since I posted properly here. Hopefully this post will make one of the reasons why obvious (there is a small postscript at the end as to the other reason).

Today, the Verge has published a lengthy, reported feature I’ve worked on for most of this year about the so-called “censorship-industrial complex” – the US right’s name for mis- and disinformation researchers across the world – and the backlash against it in Donald Trump’s second presidential term.

It’s long, but hopefully an interesting read, for which I travelled to DC, attended multiple Congressional hearings, did dozens of interviews, and read tens of thousands of pages.

Here’s a small extract to whet the appetite:

The day’s star witness was Matt Taibbi, an independent journalist and onetime liberal darling, who had been among those people handpicked by Elon Musk to publish revelations from the so-called Twitter Files, exposing — as they saw it — how concerns about “misinformation” had been exploited to censor conservative and dissenting voices on the platform.

Taibbi and company were calling for the government to do more in the name of free speech — defunding any efforts funding fact-checkers or misinformation research, and similarly ending US government funding of media across the world, which they dismiss as “propaganda.” Over the last few years, Musk, Jordan, and Taibbi had created something of an unstoppable machine: Jordan had the power to subpoena evidence, call witnesses, and create reports. Taibbi and others could testify at those hearings and report on them, as well as on material provided by Musk. Musk, in turn, could launch lawsuits based on the findings of Jordan’s committees and on the reporting of Taibbi and others.

To those people caught in that machine, though, things looked very different. From their perspective, they had been trying to protect America’s free speech. During the heights of covid, false information that stopped people from getting vaccinated or from masking, or which made them try unsafe “cures,” could prove fatal. The January 6th protests showed that political misinformation could be a life-and-death matter, too.

And now, the people who had tried to force social networks to take these issues seriously found themselves condemned in Congress, blazoned across Fox News, facing death threats and the end of their professional careers.

What started with a row over fact-checking and moderation of particular stories on social media — the Hunter Biden laptop, the Wuhan lab leak theory of covid, the QAnon conspiracy theory — has turned into a worldwide battle on the nature and limits of free speech online, covering anywhere and everywhere the government interacts with social media companies, or where it funds anything relating to media. Even the future of the transatlantic alliance is at stake after JD Vance accused Europe of becoming an enemy to free speech.

But at its core, this is still a bloody fight over what is and isn’t true — with claims and counter-claims thrown in every direction. At various points, people involved have accused one another of being former CIA spies or PR flacks for Hugo Chávez, of having flung a custard pie laden with horse semen into the face of a rival, and more. (Almost all of the above turned out to be — more or less — true.)

Please do read the whole thing over at The Verge. Please. I spent so long working on it.

Anyway. What, dear reader, does that have to do with us over here? Surely this post isn’t just shameless self-promotion? Happily, I can say that it isn’t. Matt Taibbi, the intrepid “Twitter Files” adventurer named above, declined to be interviewed for the feature, but he did provide a lengthy response to some of the points put to him.

Even in a piece that weighs in at more than 6,000 words, it was far too long to include in full, and as you’re about to see – and as the feature discusses – quickly gets bogged down in detail that is almost impossible to understand unless you’ve spent months of your life looking into it all.

The Verge feature includes a flavour (I did 6,000 words of US spelling over there, you get Britishisms here) of Taibbi’s response, but in the interests of fairness I thought it only right to include it in full here. I’ve interspersed it with a bit of commentary from myself, and occasionally from Renée DiResta, a major subject of the feature and the former research manager at the Stanford Internet Observatory.

So, here’s the correspondence in full, starting with my opening request for an interview:

Date: Wed, 14 May 2025 22:27:30 +0100

From: James Ball <REDACTED>

To: Matthew Taibbi <REDACTED>

Subject: Chat for The Verge?

Matt,

Hope this finds you well. I’m working on a feature for The Verge on the ongoing battle between disinformation researchers / the censorship industrial complex and their critics – not least the House Judiciary Committee and the current administration.

Obviously you and your work have been major players in all of this and I’d love to have a chat (over Zoom or similar?) if you had time in the next couple of days.

I’d be looking to touch on your work on the Twitter Files, your evidence to the House Judiciary Committee (particularly the hearings this year) and some of the activities of the administration with connection to USAID-funded newsrooms and similar.

Do let me know when might suit, or if you’ve got any questions etc.

All the best,

James

The first response was, ah, brief:

Date: Wed, 14 May 2025 23:22:01 +0000 (UTC)

From: Matthew Taibbi <REDACTED>

To: James Ball <REDACTED>

Subject: Re: Chat for The Verge?

James,

Thanks, but no thanks.
Sincerely,
Matt

Leading to this response, in which I give Taibbi the chance to respond to possible criticism that may appear in the final article:

Date: Fri, 16 May 2025 14:00:00 +0100

From: James Ball <REDACTED>

To: Matthew Taibbi <REDACTED>

Subject: Re: Chat for The Verge?

Matt,

No problem, totally understand – given your long involvement in the topic, you’re likely to feature in the article in any case, no in the spirit of ‘no surprises’ I’d like to give you a chance to respond to the following points, which have come up either from other interviewees, your testimony to public hearings, or other research for the article.

Any response you have to any of the bullets would be much appreciated, as well as any generalised comment you’d like to make.

Background/historical claims

• Your book “The eXile: Sex, Drugs, and Libel in the New Russia”, published as memoir, details behaviour that would now be regarded as harassment and abuse of women.

• In response to these allegations being made publicly, based on that book, you have offered an apology, but also claimed that much of that book is fictionalised.

• You have filed a defamation suit against Democratic congresswoman Sydney Kamlager-Dove in connection to a tweet she made in connections to her own comments at a hearing calling you a “serial sexual harasser”.

• The book also recounts an incident in which you hit a New York Times journalist with a custard pie laced with horse semen

Twitter Files and related claims

• An early Twitter Files story you reported concerned Twitter receiving payments from the FBI. This was widely interpreted as relating to payments for content moderation or censorship, but in reality concerned payments for cost of compliance with law enforcement requests. These payments had been previously publicly disclosed

• Another early story concerned the Biden campaign – prior to his involvement in the White House – asking for certain posts related to the Hunter Biden laptop to be deleted. These tweets turned out to contain non-consensual nudity, against Twitter ToS.

• You have repeatedly accused the Stanford Internet Observatory of working to censor 22 million tweets. This figure instead relates to 22 million tweets identified after the time as being flagged for misinformation. The true figure of flagged tweets is instead around 3,000.

Evidence to Congressional Committees

• In evidence to Congress, you repeatedly confused the timetable over when the Disinformation Governance Board was launched and then shuttered, saying it took place in 2020, leading to the establishment of other means of censorship. In reality, the DGB was launched and shut down after the other bodies were established

• You have claimed on multiple occasions that the statistic that posts connected to the Internet Research Agency were viewed by 126 million people during election originated with Renee DiResta. In reality, it was contained in evidence given by Meta to Congress. This claim can be quickly and easily publicly verified.

• Your evidence on multiple occasions confuses the government body CISA (the Cybersecurity and Infrastructure Security Agency) with the non-governmental body CIS (the Center for Internet Security). This confusion leads to the false claim the former body was making content takedown requests that actually originated from the latter.

• You likened news organisations funded by USAID or Internews to a “rot” and a corrupt propaganda mechanism. The defunding of such newsrooms at short notice has led to Russian journalists in exile losing their visas at short notice, and to police raids in at least one newsroom in Serbia.

Sorry for the relatively long email, but did want to give you a chance to respond to any potentially negative comments the piece may contain, even if you were denying an interview.

All the best,

James

And now the big one:

Date: Fri, 16 May 2025 11:20:03 -0400

From: Matthew Taibbi <REDACTED>

To: James Ball <REDACTED>

Subject: Re: Chat for The Verge?

James,

Well, I can see where you’re going, and with whom you’ve been talking. I hope you’re proud of yourself.

First of all, I think you have me confused with Michael Shellenberger on a few occasions, but we’ll leave that aside. I hope you’re going to be careful to use my actual words.

Second, I categorically deny that I’ve ever, in Russia or anywhere else, been involved in behavior that could be considered harassment or abuse of women. No woman has ever accused me of any kind of sexual impropriety. The notion that there was “abuse” or “harassment” is based entirely on one passage written by Mark Ames that purports to show me not abusing anyone but laughing at something, in a scene he himself describes as fictional (it is a fictionalized take on a real scene in which I did say “But it is funny” when a humor-challenged co-worker complained something wasn’t funny – but that thing was not sexual harassment). The actual women described in the passage have been interviewed and (as multiple publications with whom I’ve reached financial settlements - make sure your editor takes note — have been forced to concede) and said they were never harassed by me. See for instance https://www.theguardian.com/commentisfree/2017/oct/19/liberal-men-feminism-harvey-weinstein.

I’m suing Rep. Kamlager-Dove because I’m tired of people publicly issuing this damaging accusation of harassment or sexual misconduct while simultaneously showing a total lack of interest in whether or not this ever occurred. The original reporter on that story, when asked why she didn’t attempt to contact my co-workers, said, “I have not written about these accusations as a journalist.” You may want to try calling anyone who’s ever worked with me to see if I’m a “serial sexual harasser” or a father of three married for nearly fifteen years about whom there is not and has never been even a hint of a rumor of sexual impropriety.

Now for just a few of your issues:

* Your evidence on multiple occasions confuses the government body CISA (the Cybersecurity and Infrastructure Security Agency) with the non-governmental body CIS (the Center for Internet Security). This confusion leads to the false claim the former body was making content takedown requests that actually originated from the latter.

As I’ve repeatedly explained, the Center for Internet Security, or CIS, and the Cybersecurity and Infrastructure Security Agency, or CISA, were both parties to the Election Integrity Partnership. EIP has plainly admitted this. You can find that here https://www.eipartnership.net/blog/a-statement-from-the-election-integrity-partnership or look on the “Four Major Stakeholders” entry here, on page 12https://stacks.stanford.edu/file/druid:tr171zs0069/EIP-Final-Report.pdf. In that same list of “Four Major Stakeholders” you’ll see something they call EI-ISAC, or the Elections Infrastructure Information Sharing & Analysis Center. That is CIS: https://www.cisecurity.org/ei-isac. Complicating matters, CIS was significantly (and according to my sources, almost exclusively that year) funded by CISA:

https://www.usaspending.gov/award/ASST_NON_19PDMSI00002_7061

Just interrupting here to flag that I’m not sure what Taibbi thinks this previous paragraph demonstrated. The EI-ISAC initiative he mentions is indeed CIS, which doesn’t do anything to clear up the apparent (repeated) confusion between CIS and CISA in Taibbi’s testimony.

During the 2020 election, when outside bodies wanted to file “misinformation” complaints to the EIP portal (a typical example would be the Georgia Secretary of State’s office upset about a tweet with inaccurate poll closing times), they would send the complaint to misinformation@cisecurity.org, i.e. to EI-ISAC. Once sent, the complainants would receive a form notice back from CIS that read:

“We have already forwarded your report to our partners:

1. The Cyber and Infrastructure Security Agency at the Department of

Homeland Security. They will submit it to the relevant social media

platform(s) for review.

2. The Election Integrity Partnership”

In. other words, the complaint split in two. One copy went to EIP. The other copy went to CISA, which in turn appeared to forward complaints to the platforms like Twitter, where we would see them. The EIP complaints that we looked at that arrived in Twitter’s JIRA ticketing system were typically identified by what looked like a CIS complaint number (for instance, CIS-MIS000230), but the actual complaint had not come from CIS. It went from the Original Source to CIS to CISA to Twitter. On the last step of the journey, it would in fact usually be forwarded from a DHS account (e.g. Some.Guy@cisa.dhs.gov). So we’d be left with something that would say CIS-MIS10012 in the subject line, but came from CISA, along with yet another form notice that read, “The Cybersecurity and Infrastructure Security Agency (CISA) of the U.S. Department of Homeland Security (DHS) is not the originator of this information. CISA is forwarding this information, unedited, from its originating source.” This is a sample of what complaints we saw in the Twitter Files looked like:

Hi GETSupport, can you please review the report below from the state of Washington.

---------- Forwarded message ---------

From: Stafford, John <John.Stafford@cisa.dhs.gov>

Date: Wed, Oct 28, 2020 at 11:46 AM

Subject: FW: Case #CIS-MIS000079: False poll numbers to boost write-in candidate’s campaign

To: scardille@twitter.com < scardille@twitter.com>

Cc: CFITF All < CFITFAll@hq.dhs.gov>

Good morning, Please see below reporting from Washington state. Best,John John Stafford

Analyst, Countering Foreign Influence Task Force

National Risk Management Center (NRMC)

Cybersecurity and Infrastructure Security Agency

U.S. Department of Homeland Security

(o) (703) 603-4969

(c) (202) 510-0397

john.stafford@cisa.dhs.gov

Two things worth noting here. The first is that both CIS and CISA could, and did, flag content to Twitter directly without involving the EIP. The fact they did so seems to me to greatly weaken the argument that the EIP was either created as or used as a carve out. If they are reporting directly anyway, what would they gain by also having a workaround they don’t always use?

The second is that the above email, according to DiResta, “did not involve EIP at all”, because “EIP did not assign CIS numbers”.

When I did that Twitter thread mentioning the Slack notation “according to CIS (escalated via EIP)” I genuinely thought the Twitter employee was referring to CISA and had simply mistyped, because I had only seen hundreds of communications of this type that came to Twitter from CISA. I wrote “CIS(A)” to make it plain that I was adding the A. Mehdi Hasan said I’d made a “false claim” that EIP was partnered with the government, which was a far more serious error than my own – EIP by its own admission admitted was partnered with CISA, and in fact emails were later found in the Twitter Files (and via Jim Jordan’s Committee investigation) strongly indicating EIP was founded at the behest of CISA. “We just set up an election integrity partnership at the request of DHS/CISA” was just one of the emails Jordan found, from one of the EIP partners. Many others you can find here. ¹

This is getting to the core of the row between Taibbi and DiResta, so I’m going to include some rebuttal from DiResta directly here²:

EIP chose to escalate some tips that came to us from CIS after doing our own independent analysis and determining that they were worth escalating; we chose to ignore far more. That is independent research. We didn’t receive any requests from the government, or CIS, to ask platforms to do anything. Matt hasn’t provided any emails to that effect because none exist.

Matt’s tweet said: “It’s crucial to reiterate: EIP was partnered with state entities like CISA and GEC while seeking elimination of millions of tweets. In the #TwitterFIles, Twitter execs did not distinguish between organizations, using phrases like “According to CIS[A], escalated via EIP.”

This is just nonsense … As for Matt’s blog post and the email he references: CISA did not found, fund, or otherwise control the EIP … This was all clarified in testimony excerpts from Graham, Alex, & Kate released by Congress minority here.

Additionally, there was no government funding until the NSF grant two years later.

Back to Taibbi:

Jordan also produced an email (it’s in the above article and this report) from Facebook summarizing a conversation with DHS in which DHS officials reportedly said “DHS cannot openly endorse the portal” (i.e. EIP) but wanted “behind-the-scenes” access and “would like to have incoming the same time the platforms do.” This would explain the bizarre ““We have already forwarded your report to our partners” structure in which EIP and CISA were both copied on complaints. In hindsight I should have just let it be, but CIS/CISA was a very complicated issue, in which CIS appeared to play no role beyond acting as a middleman between agencies. The notion that DHS/CISA was not involved with sending complaints is absurd, belied by countless communications we’ve published, including an email from Elvis Chan to Twitter saying the FBI would be the “belly button” for USG complaints, “We can give you everything we’re seeing from the FBI and USIC agencies. CISA will know what’s going on in each state.”

DiResta once again has strong pushback here:

Matt is once again misunderstanding a significant thing: In these exchanges, CIS was discussing creating its own portal. These emails are from Jan 2020, months prior to the idea of EIP. Jordan’s report, which Matt references, describes “a “Misinformation Reporting Portal” to be operated by the Center for Internet Security (CIS), a non-profit funded in part by CISA.” This is on p 23. EIP_Jira-Ticket-Staff-Report-11-7-23-Clean.pdf The planning exchanges between tech companies and government and CIS in these emails have nothing to do with EIP.

The complexity of these exchanges, incidentally, is a large theme of the article. Everything is acronym-dense and reliant on snippets of much larger ongoing exchanges between colleagues, meaning the way they’re read depends on how they are framed to a new audience. Time and again in the reporting of the article, a seemingly damning quote would turn out to be much more dull.

There is very obvious a legitimate question around the extent of roles of NGOs in moderation and content flagging – especially when they cooperate with tech companies and perhaps governments – but time and again the most obvious bombshells would turn out to fizzle once you looked at them closely.

Anyway, back to Matt:

As for the claim that the EIP only flagged 2,890 tweets instead of 22 million, first of all, you’ve got it written not quite right here: “The true figure of flagged tweets is instead around 3,000.” The language they actually used was “unique tweet URLs.” https://cyber.fsi.stanford.edu/io/news/background-sios-projects-social-media

My wording – “According to the EIP’s own data, it succeeded in getting nearly 22 million tweets labeled in the runup to the 2020 vote” – was based on a line from the EIP’s own final report:

“In total, our incident-related tweet data included 5,888,771 tweets and retweets from ticket status IDs directly, 1,094,115 tweets and retweets collected first from ticket URLs, and 14,914,478 from keyword searches, for a total of 21,897,364 tweets.”

The EIP may have only “flagged” a certain number of tweets, or individual URLS belonging to problematic tweets, but each of those individual URLs about individual topics like “Sharpiegate” or “Minnesota ballot harvesting” according to their own report could correspond to many more than one “incident,” meaning people retweeting or spreading the “problematic” content. Just the category “Dominion Voting Systems : SwingStates” led to 7,157,398 “incidents,” as you can see on page 285 of their report. “Stop the Steal” spurred more than 2.8 million “incidents.” They describe the process of “collapsing” many incidents to one ticket, for the sake of “tractability”:

“In total, the EIP processed 639 unique tickets and recorded 4,784 unique original URLs. After our real-time analysis phase ended on November 30, 2020, we grouped tickets into incidents and narratives. We define an incident as an information cascade related to a specific information event. Often, one incident is equivalent to one ticket, but in some cases a small number of tickets mapped to the same information cascade, and we collapsed them. As described in Chapter 3, incidents were then mapped to narratives—the stories that develop around these incidents—where some narratives might include several different incidents. For tractability, we limited our analysis in this chapter to 181 tickets mapped onto 153 incidents related to the narratives in Chapter 3 and that we determined to either (1) have relatively large spread (>1000 tweets) on Twitter, or (2) be of “high priority” as determined by analysts during our real-time research.”

This is a central claim and is dealt with at some length in the article. It still seems to me that Taibbi has misread or misunderstood what the report (which is available online and linked from the piece) actually says.

DiResta’s rebuttal here is also fairly lengthy, but it’s a crucial point:

That is not what the report describes. This is backwards. If you look at the table on p 185 it has a table with Incident Titles and Descriptions. Dominion Voting Systems was the incident. Description: “This incident accused Dominion Voting Systems software of switching votes in favor of Joe Biden, particularly in swing states like Georgia; as of January 2021, Dominion has filed defamation lawsuits against prominent individuals and media that perpetuated this claim, and some have retracted their stories”

The column, # of Related Tweets, 7,15,398, is how many tweets were related to the incident. When you add those up - the top 10 most viral incidents - you get 22 million.

We did not describe the process of “collapsing many incidents to one ticket at all” - it went the other direction. We grouped tickets - which student analysts created in realtime on election day - into incidents after the election. This is explained on p 49, in the Incidents and Narratives chapter (which follows a chapter on Tickets and summary statistics!): “The work of narrative identification began on November 30, 2020, after the EIP’s monitoring mission had concluded. We first grouped tickets into “information cascades,” or incidents, tracing how a single real-world event (like a video of poll workers collecting ballots in California) could generate a number of different false claims, spread at different rates on different platforms by different actors.”

This bizarre theory doesn’t even make sense - all of the people retweeting a tweet are still sharing one URL. You can see the JIRA tickets, again, they are public. This convoluted story is what Benz and Taibbi tried to come up with to save face – they need the “22 million censored tweets” story to be true somehow, otherwise the “Censorship Industrial Complex” house of cards collapses pretty hard.

· An EIP video put out in 2021 echoed the report, noting that “35% of the URLs flagged were actioned under remove, reduce, or inform” policies. That would have meant a sizable number of posts, given that each “misinformative” URL that made it through the ticketing system often appeared in Slack chats or emails tied to much larger numbers of individual tweets. In hindsight, I should have just have written something like “EIP identified nearly 22 million tweets out of 859 million as false or misleading.” It’s something I regret. But while it’s been claimed I confused the “post-hoc” research of historical tweets of the EIP with its “real-time” flagging activity, it’s important to remember the EIP and CISA have both been extremely unreliable when it comes to the scope of the “real-time” activity. Statements like “CISA did not found, fund, or otherwise control the EIP” (see above: there’s strong evidence they founded and at least partially funded the program, through CIS for instance) and “EIP did not make recommendations to the platforms about what actions they should take” (Jordan’s committee eventually found, after EIP was forced to answer a subpoena, dozens of concrete recommendations e.g. “We repeat our recommendation that this account be suspended for the duration of election day from posting additional misleading information about voting”) were clearly false. I think my efforts to describe EIP/CISA’s election work, given the material I had to work with and the abject refusal of sites like yours to investigate, has held up. If the injustice was so great, I’m surprised the SIO is not still in operation.

I could go on, but obviously it won’t help. Good luck with your bootlicking bullshit hit-piece of a story. I’m sure you’ll be editing The Atlantic in no time.

Sincerely,

Matt

Phew. Almost there. One last general point from DiResta:

The EIP’s work was First Amendment protected speech. The EIP publishing findings were First Amendment protected speech. Students and researchers sending opinions on social media posts to platforms was First Amendment protected speech. Students and researchers sharing their opinions with the government is First Amendment protected speech. The use of Congressional and Judicial power to punish this research is the real violation of free expression ideals.

As to the last point: Taibbi’s repeated misrepresentations about EIP and SIO were the pretext that enabled the use of government power to suppress the First Amendment protected expression of Stanford’s students and researchers.

Let me stress: I am aware that this post on its own is terrible content. But I hope for those who have read the original feature it serves as an interesting postscript, a living example of some of the issues it highlights – and of the day-to-day lives of some of the people caught up in the maelstrom.

Or perhaps it’s a “bootlicking bullshit hit-piece of a story”. It is, despite the best efforts of censors everywhere, still a free country. You get to decide for yourself!

---

Finally, two housekeeping bits. The first is that this is my side Substack. You don’t get any extra content if you pay to support it, but you do help financially support the side projects I do here, including this hideous 24/7 AI-generated US politics podcast, which I still host. Please do help me lose slightly less money each month funding that travesty.

And a free subscription just keeps you across what I do here, which sometimes is mind-numbingly detailed email correspondence! The next piece coming here, though, is the long and surprisingly painful saga of what happened when I tried to set up an AI model to be my subeditor, which had so many twists and turns that I’ve written and deleted it three times already. Can we get it out before 2026? Let’s find out!

1. This was a URL not a hyperlink in the original, but it was rendering as a Substack box and breaking the blockquote format, so I changed it slightly. Just ‘fessing up here for transparency

2. Sorry this is probably horrible formatting but I needed it to look different from Taibbi’s email