Our weekly Briefing is free, but you should upgrade to access all of our reporting, resources, and a monthly workshop.
This week on Indicator
Craig published a guide to what still works for Reddit search, monitoring, and archives. Reddit OSINT has changed dramatically over the past three years; this piece gets you caught up. (It includes Reddit’s recent decision to end support for RSS feeds…)
Alexios found that Google is failing to uphold its promise to label ads placed by unofficial websites that sell overpriced government documents such as driver’s licenses, passports, and visas. Of more than 2,000 ads reviewed, only 8% carried the mandated disclosure.
We also launched the Indicator Lab! It’s new membership that gives you a powerful suite of five tools plus expert guidance to run AI-assisted investigations. Background a company, trace a crypto address, scrape a website, track a ship — you can save time and multiply your findings thanks to powerful agents, 24/7 monitoring tools, and access to hands-on support. Learn more.
Grokipedia is back (kinda)
I got a message on Wednesday from researcher and friend-of-Indicator Renée DiResta that Grokipedia was accepting edits again after a hiatus that lasted more than five months. This appears to be part of a broader update to Musk’s AI-written encyclopedia that includes a more colorful homepage.
But don’t get too excited by the relaunch — Muskipedia is still struggling to stay current. SpaceX’s page mentions its June 2026 IPO in the third paragraph, but reverts to referring to it in the future in subsequent sections.
Similarly, the Indian Premier League’s entry highlights the Royal Challengers Bengaluru’s 2026 title win. But the 2026 season is absent from the “Overview of all seasons” section, which on Friday morning claimed that “the tournament [is] scheduled to begin on March 28, 2026.”
Several pages appear to be stuck in the past. I reviewed the Grokipedia pages for 10 countries that saw a change in leadership since the encyclopedia’s edit freeze in April. Eight pages list the wrong head of government. On Grokipedia, for example, Keir Starmer is still Prime Minister of the United Kingdom. Andy Burnham has been leading the country since July. (The relevant update on Wikipedia was made on the same day.)
Benin’s Grokipedia entry lists Patrice Talon as president even though he was replaced by Romuald Wadagni in May.
In the Solomon Islands, Jeremiah Manele was booted out by a no-confidence vote and replaced by Matthew Wale in May. But Grokipedia says Manele is still in charge.
The pages for Colombia, Costa Rica, Iraq, and Morocco don’t even mention the new leaders that took power or were designated between May and September of this year.
Hungary is a little better. Péter Magyar’s victory over Viktor Orbán is reflected in the text, but his premiership is only referred to in the future tense.
All this to say: I would probably avoid Grokipedia if you’re preparing for trivia night — Alexios
Deception in the News
📍 Adam Schiff and Alex Padilla, the Democratic senators for California, wrote a letter to the heads of Google, Instagram, Meta, Snap, SpaceXAI, TikTok USDS, and YouTube asking for information about the platforms’ plans to combat disinformation about the 2026 midterm elections. The questions (below) are pretty detailed. We’ll see about the answers, which are expected by Oct. 15 — Alexios.
📍 A Wesleyan Media Project review of 164 AI-generated or enhanced ads placed during the US midterms and found that 69% did not disclose the use of AI. Labels were often absent in states where the disclosure is legally mandated.
📍 Maldita.es CEO Clara Jiménez Cruz asked French President Emmanuel Macron about the EU Digital Services Act at a dialogue in the Complutense University of Madrid. Macron responded that there’s a lot more to do to enforce the DSA and that trusted third-party fact-checkers are essential.
📍 Proofpoint said that China-linked hackers impersonated former US government officials and an Anthropic employee in an attempt to extract sensitive information about AI development from US policy experts.
📍 Italian media reports that Paolo Molesini, then-president of Fideuram bank, transferred 95 million euros to scammers as a result of AI-generated audio and spoofed messages. The scammers impersonated Carlo Messina, CEO of Intesa Sanpaolo, a major banking group, and Paolo Nastasi, the managing partner of a law firm. Molesini knew Nastasi well and still fell for it.
📍 An Alaskan woman self-reported to local wildlife authorities after hunting snipe out of season — a mistake she blamed on a hallucinated Google AI Overview answer. According to an affidavit reported by Alaska Beacon, the woman “stated that she Googled ‘Snipe season’ on her phone and that, according to Google, snipe season began on September 1st.”
📍 In more AI-related trouble, a Singaporean man was charged for sharing a synthetic image showing a crocodile in Pandan Reservoir. BBC reports that “the national water agency suspended its work at the reservoir for two days,” due to the image. The man “could face a prison sentence and a large fine if he is found guilty.”
Tools & Tips
Good news: Google Earth added a “Detect change” tool that compares two years of imagery and highlights changes.
Select the tool, draw a polygon around an area, and choose two years to compare. As in the above example, Google Earth visually highlights areas where changes have occurred. The deeper the yellow shading, the more significant the differences.
“This tool will really tell us what has changed in the area,” said Juan Francisco Saldarriaga, a senior computational designer who works on Google Earth, in a video demo.
“It’s something we #investigators and journalists do all the time, and there are plenty of crisp, high-resolution images that could make this genuinely useful for that kind of work,” wrote journalist Ben Heubl on LinkedIn. — Craig
📍 Four New York Times staffers authored a research paper about how they developed an agentic search tool to allow reporters to dig into the Epstein Files. “The Engine translated reporter questions into Google BigQuery SQL queries across three corpora: Epstein-related releases, the Times's archive and external, Epstein-related news headlines,” they wrote. “It used an LLM to plan queries and returned citation-rich answers a reporter could verify and trust.”
A notable passage:
A newsroom agent is most useful as an interface to institutional memory and source material, not an autonomous writer. Its value lies not in knowing an authoritative, final answer, but in helping journalists pose better questions and weaving evidence across data silos into answers they can independently verify. It grants them analytical capabilities previously unimagined and then leaves judgment in their hands.
The NYT project is also described in this slide deck about recent Epstein Files-related projects in newsrooms. — Craig
📍 Lilith Wittmann, an independent journalist in Germany, published curacaofiles.com, a site that allows you to search in “over 40,000 documents of casinos licensed by the Curaçao Gaming Authority.” The documents were collected as part of the #CasinoSecrets series she worked on with media partners.
📍 Nicole Hurey of The OSINT Vault launched a free tool called TELLTALE. “It examines writing for characteristics associated with AI-assisted writing, and shows you the evidence behind each one: what was found, where, and what it does and does not establish,” according to the site. It looks for zero-width characters and non-breaking spaces, among other potential indicators of AI-generated text.
📍 Two new tools on GitHub I haven’t tested yet (always be cautious when running other people’s code!):
Patrick Swanson of Verso released Big if True, a fact-checking skill for Claude. From the website: “It checks your draft one claim at a time and shows you what needs a second look, with sources.”
Geo-sleuth is a free tool that “finds where a photo was taken — OpenStreetMap geometry, elevation skylines, satellite imagery and street view — and shows its work.”
📍 MJ Banias wrote, “A Free OSINT Lesson: Fashion, Optics, and Paper Trails.”
📍 Benjamin Strick released a new video about a Russian base “hidden beneath a fake forest on satellite maps.” It’s great case study and a helpful tutorial for “how to identify what an image shows, confirm its location, check when it was captured, and assess whether it supports the claims being made about it.”
Reports & Research
📍 Nature found that sham scientific societies are doling out honorific titles to established researchers in order to project legitimacy and charge membership fees to more junior scientists. Particularly galling to me was a Google Maps location pinned on the European Academy of Engineering’s website that “displays an image of a stately academic building with large block letters on its facade that say European Academy of Engineering in Swedish.” In reality, it is an altered image of the nearby University of Gothenburg. — Alexios
📍 A new paper in the Journal of Online Trust and Safety analyzed the companies providing infrastructure to 88 websites hosting nonconsensual intimate imagery (including of the synthetic form). The paper call out five players in particular: “Cloudflare, Google, Namecheap, WordPress, and Proton Mail.”
📍 Russian influence operations that create fake news reports mimicking real Western outlets continue to churn out content, according to a NewsGuard report. It identified 421 fabricated stories in the first half of 2026. Since Russia's full-scale invasion of Ukraine in 2022, the campaigns have mimicked 163 media outlets in 906 fake reports.
📍 WHAT TO FIX released a report about “how weaknesses in Meta’s monetization systems and governance may expose children to harm.”
📍 Reporters Christo Grozev and Tatsiana Ashurkevich hired someone on a Russian marketplace to create a fake video showing them meeting with a NATO general on a yacht. They paid a placement broker about £60 more to have it published as news. Within an hour, it appeared on six Russian websites. Grozev broke it down in a video that details “how the Kremlin is using new digital tools—including generative AI—to reshape propaganda for the algorithmic age.”
From a description of the investigation:
Rather than relying only on traditional media or trolls, Russian operators now blend fake-news sites, manipulated videos, paid influencers, and AI-generated content into one coordinated system of influence. AI-created “experts” and synthetic witnesses appear alongside genuine commentators, helping Kremlin narratives reach Western audiences through familiar formats and credible platforms.
One More Thing
Ecommerce lawyer Rob Freund, whom we’ve quoted in stories about undisclosed ads, often shares examples of cases where lawyers have been caught citing AI-hallucinated cases in court filings. He recently tweeted out a new milestone: “a lawyer admitted ChatGPT wrote the briefs; made up cases, people, and testimony; and he didn't check any of it.”
Here’s the order from the judge:
Indicator is a reader-funded publication.
Please upgrade to access all of our content, including our how-to guides and Academic Library, and to our live monthly workshops.
