🤥 Faked Up #10

Negated Notes | Dissing RT | Unlabeled

Faked Up #10 is brought to you by AI-generated barndominiums and facts that are facts. The newsletter is a ~6-minute read and contains 43 links. I will be on holiday July 15-26 but there’s a special edition of Faked Up dropping next Wednesday reviewing three new books about digital deception. Regular programming resumes July 31.

Top Stories

NEGATED NOTES — Clashes among contributors to X’s Community Notes are leading to fact checks being appended late or not at all on misleading tweets, according to Logically Facts.

Take the notes below a misleading post accusing Congress leader Rahul Gandhi of calling all Hindus violent. Because some raters claimed NNN (no note needed), the post lacks a visible fact check 9 days after being posted.

X’s requirement for cross-partisan agreement on notes can in theory act as an incentive for high-quality sourcing. But blocking a label from appearing introduces a juicy incentive for counter partisan brigading.

There are signs this may not be a one-off.

Mediawise director Alex Mahadevan, who is my go-to Community Notes connoisseur, told Logically that fewer than 10% of notes go live. That share shrinks to 4% for particularly controversial topics like the Israel-Hamas war. Mahadevan said Community Notes’ “biggest weakness is that it requires bipartisan agreement on facts. And that's just not how facts work, a fact is a fact. A number is a number.”

ANGRY CLUSTERS — A group of researchers at the Zhejiang University of Technology claim in a preprint that Reddit threads about false claims tend to have more back-and-forth and be more negative in tone than those about true claims. (They based their analysis on a previously published dataset of Reddit posts tied to fact checks by PolitiFact, Snopes and Emergent.info.) Not sure if this is a terribly novel finding, but this chart is pretty.

DISSING RT — Journalists at The Insider1 and Der Spiegel got their hands on correspondence between officers in the SVR, Russia’s foreign intelligence agency. The documents describe a proposed influence operation nicknamed Project Kylo aimed at changing Western perceptions of the war in Ukraine.

There’s a lot in these exchanges that corroborates prior findings that Kremlin-backed influence operations seek to foment existing divisions in the West. For instance, a 2022 memo suggested pushing stridently pro-Ukrainian demands to elicit a backlash:

“Waging network wars in EU cyberspace based on the increasing demands of Ukrainian migrants and the new waves of irritation of the local population provoked by this, according to preliminary estimates, will have a very high efficiency both now and in the foreseeable future.”

What caught my eye, however, was the denigratory tone used in the proposal to describe RT, Sputnik and other “old” state-controlled media outlets. SVR officer Mikhail Kolesov claims that these organs “have demonstrated near-zero effectiveness for decades, not years.” This should certainly be interpreted in part as the bluster of someone trying to pitch an alternative approach, but it is a good reminder that Russian state propaganda is not a unitary block.

Incidentally, RT does appear to be innovating in its propaganda operations. According to Canadian, Dutch and US intelligence agencies, RT affiliates have been using a tool called Meliorator to create inauthentic X accounts amplifying pro-Kremlin content.

IN ADS WE DISTRUST — In a preprint whose study design I can only define as chaotic good, two computer scientists asked 23 local politicians from Germany and 20 journalists from the US to describe how they would rate three low-credibility news sources. The researchers found there were 11 criteria that respondents tended to use to assess the trustworthiness of the site in front of them. These included the website’s content, reputation and self-description.

While the two groups mentioned some criteria at similar rates (see chart below, where “experts” are the journalists and “laypeople” are the elected officials), two exceptions stand out.

18 out of 20 journalists looked for the author of an article and considered their biography. Only 9 out of 23 politicians did the same.

Equally interesting was the differential reliance on Ads as a proxy for credibility. This was the single-most cited (negatively impacting) factor for politicians but only came up from 4 out of 20 of the journalists. I wonder whether this is because the latter are more familiar with the reality of online business models for news and more inured to the dissonance of seeing a lousy ad next to a credible news article.

LYIN’ CRIES — “Marine Le Pen crying” was trending after the far-right leader underperformed in the French legislative elections.

This was thanks to posts like this by the 100% non-bot-passing X account “@MrK55082208.” The video actually shows Le Pen crying because she was laughing vigorously. The clip is from a 2021 video, Le Pen was listening to a satirical impersonation.

UNLABELED — Speaking of the French far right: AI Forensics claims that the Reconquête and Rassemblement National parties were by far the most frequent posters of AI-generated images on social media during the country’s EU and parliamentary elections. Many of these images are used to support a slogan rather than to tell a free-standing story (see below).

In this sense, they are more akin to stock photography — versatile and often clichéd imagery used to make a message pop — than to manipulated images. And that’s how the parties appear to be treating them too, with one RN deputy telling Politico that “When you compare the price of a Shutterstock subscription and a Midjourney subscription, Shutterstock becomes irrelevant.”

The bigger story for me is that none of these stories are labeled by either political parties or the platforms. There are reasons for this that involve the transparency standards Midjourney has not signed on to. But I can’t shake off the feeling that watermarking, a popular mitigation strategy for AI deception on paper, remains rather pathetic in practice.

SYNTHETIC NCII CORNER — There is unfortunately enough reporting on non consensual deepfake porn to fill a dedicated newsletter. Here’s three recent items:

• Channel 4 found that at least twelve British female politicians have been undressed with AI nudifier apps. They include Labour’s Deputy Leader Angela Rayner, Education Secretary Gillian Keegan, Conservative Commons Leader Penny Mordaunt, former home secretary Priti Patel and Labour backbencher Stella Creasy.
• NBC News claims that Mastercard and Visa cards can still be used to buy nonconsensual deepfake nudes on Fan-Topia through a hidden links intermediary, despite the credit card companies prohibiting this use of their platforms.

The Balkan Investigative Reporting Network uncovered at least 20 active Telegram channels for AI undresser bots, six of which appeared to be tied to a user named “Dzoni.”

Headlines

1. The $11 Billion Marketplace Enabling the Crypto Scam Economy (Wired)
2. Victor Miller confirmed as mayoral candidate for Cheyenne; AI program VIC excluded from ballot (Cap City News)
3. Two years after an open letter to YouTube, fact-checkers remain dissatisfied with the platform’s inaction (Poynter) with ‘Frenemies’: The complicated relationship between fact-checkers and tech giants like Meta and TikTok (Poynter)
4. Real criminals, fake victims: how chatbots are being deployed in the global fight against phone scammers (The Guardian)
5. Students Target Teachers in Group TikTok Attack, Shaking Their School (NYT)
6. Das geheime Netzwerk hinter den gefälschten Promi-Werbungen (The secret network behind the fake celebrity endorsements - SRF)
7. Google Search Ranks AI Spam Above Original Reporting in News Results (Wired)
8. As sunscreen misinformation spreads online, dermatologists face real-life impact of online trends (CBS News)
9. Shadow campaign: Global influence op targets Qatar in wartime (AFP)
10. Lie detection algorithms disrupt the social dynamics of accusation behavior (iScience) with AI lie detectors are better than humans at spotting lies (MIT Technology Review)
11. While skimming through tens of Doppelgänger ads recently approved by Meta, for a total reach over 200k users in France, Germany, Italy, and Poland over the past two weeks, I stumbled upon an interesting slip-up in one of their ads 👇 (Paul Bouchard)
12. Forscher entwickeln Textscanner, um mit KI Manipulationen zu erkennen (Researchers develop text scanners to detect manipulation using AI - MDR) with Experiments in News Bias Detection with Pre-Trained Neural Transformers (arXiv)
13. The Kremlin is rewriting Wikipedia (The Economist)
14. How Disinformation From a Russian AI Spam Farm Ended up on Top of Google Search Results (Wired)

Before you go

This “barndominium” is AI-generated but you can still pay $80 for floor plan layouts that don’t match its images (Conspirator Norteño)

I…don’t think this is a good idea?

“AI Steve” did not get elected to Parliament in Brighton, coming rock bottom in the vote count behind even Citizen Skwith.

Speaking of the UK election…

Has the word “misinformation” peaked on Wikipedia?

1 If you are struggling to open the Insider link, it’s because the site has been subjected to a massive DDOS attack. Consider trying later or using an archive site like archive.is or archive.org to access the piece.