Briefing: ChatGPT for spotting scams

Our weekly Briefing is free, but you should upgrade to access all of our reporting, resources, and a monthly workshop.

Digital Digging’s Henk van Ess wrote a guest post detailing 8 AI prompting mistakes and how to fix them.

We dropped the latest episode of Show & Tell, our podcast spotlighting analysts, journalists, and researchers studying digital deception. In this episode we spoke to Renée DiResta about Grokipedia, Community Notes, and the state of platform disclosures about influence operations.

We also updated our map tracking reported cases of AI nudifiers used in school communities worldwide. The running total is now 116 cases with 916 reported victims.

And we continue to track the share of AI-generated posts in X’s Community Notes. Last Friday it hit 67.94%, a new single-day record.

📍 OpenAI published a microsite with high-level data on how ChatGPT users ask the chatbot to verify potential scams. The company said it sees 131.4 million queries related to scams each week. France, Italy, and Spain are the countries where scam queries are most common.

As welcome as this disclosure is, I can’t help but think that there is no Google Trends-like interface for ChatGPT (or Claude or Gemini). Understanding online behavior will increasingly require deeper insight into user query patterns at scale across these platforms. I hope we get more insight soon. — Alexios

📍 Freelance science journalist Caroline Williams said an impersonator was able to get a story published in Nature under her byline, and get paid for it. The publication has retracted the article.

📍 Scammers harvested sensitive data from hundreds of Revolut customers by sending fraudulent requests from a compromised Italian government email account linked to the Prefecture of Reggio Calabria.

📍 A TikToker used facial recognition to identify a man at a store and then pretended he’d read his book, filming the interaction on his Meta glasses and posting it online. There’s even more manipulative stuff on his profile.

📍 Several US-based urologists reported that scammers created AI-generated profiles using their medical credentials to scam patients. In one case, a doctor discovered the incident after a patient called her practice to complain about adverse effects from the cure prescribed by her synthetic doppelgänger.

📍 The International Federation of the Phonographic Industry announced a new Streaming Integrity Initiative to combat streaming fraud. They will have plenty of work.

📍 The Manhattan District Attorney shut down 12 websites circulating explicit deepfakes of celebrities. Wired’s Matt Burgess reports the sites “were likely viewed millions of times” before being taken offline.

📍 404 Media, Futurism, and Rest of World are among the media outlets who received unwanted pitches from AI agents powered by a platform called iLands. Ernie Smith, also a lucky recipient, dubbed iLands “Fiverr for autonomous bots.”

📍 Forty-seven Latin American fact-checking projects urged Meta not to replace its Third-Party Fact-Checking Program with Community Notes. (The company recently expanded the latter to 16 countries in the region.)

📍 WayTrace is a free tool from Thomas Housset that scans a domain’s archives in the Wayback Machine to reveal emails, subdomains, linked social accounts, exposed directories, analytics and ad IDs, and other potentially useful information.

It runs an initial scan and provides an overview of how often the site’s been archived:

Then you choose from a list of info/indicators that it can pull from captured pages before running the full scan (which can take several minutes):

The resulting report is impressive, though it depends on your target. For example, here are the analytics IDs WayTrace extracted from captured pages in the Wayback Machine.

A useful, well-executed idea. (via The OSINT Newsletter) — Craig

📍 Another Wayback Machine tool: it has a searchable collection of “3,787,236,399 Telegram web pages using an index built from anchor text, all URL parts (file names, hosts, domains), MIME Types, HTTP Status codes and the full text of the web pages.” (via @techniquesOsint)

📍 Stephen Abbott Pugh added a new feature to OpenCheck, his free tool that lets you search for data on three million companies. It pulls from “dozens of open sources including national registers, sanctions lists, procurement data, leaks databases.” You can now monitor a company and get alerts when new data is added.

📍 Nico Dekens wrote, “10 OSINT analysis techniques to accelerate investigations.”

📍 Global Fishing Watch published, “How AIS and VMS Tracking Technologies Work: 10 Essential Questions Answered.”

📍 The Open Source Centre and Maltego published, “Mapping North Korea’s Illicit Shipping Networks with GEOINT and Maltego.”

📍 Rotem Yasur and Inbar Yasur launched FIMI Monitor, a dashboard that “brings together documented incidents, research, and reports on influence operations and disinformation in Israel and around the world.”

📍 LLM Wiki Newsroom is a free, open-source “self-evolving multi-agent newsroom that turns your documents into a cross-linked markdown wiki.”

📍 Hackers Arise published, “Open Source Intelligence (OSINT): Gathering Information on Discord Users.”

📍 The Public Tech Media Lab at the University of Wisconsin Madison shared a detailed article and video about how to geolocate white phosphorus plumes using open source tools. Guillen Torres, the Lab’s director of development, also invited people to pitch ideas for other tutorials. “The criteria we'll use to choose: must be timely, original and, above all: accessible, instructive and clear,” he said.

📍 SkopeNow is hosting the 2026 edition of its free virtual OSINT Live event on Oct 15. Speakers include Micah Hoffman and Griffin Glynn of My OSINT Training, Dr Manisha Ganguly of The Guardian (and soon to be TIME magazine), and Tom Vaillant, our partner for OSINT Navigator and the Indicator Lab.

📍 The Yale School of Public Health’s Humanitarian Research Lab is hosting a paid course, “Remote Sensing and OSINT Analysis for Human Rights and Humanitarian Applications.” It costs $5,000 USD and runs from Oct 19 to 23.

📍 The Security & Intelligence Professional Studies Initiative is holding a free webinar on Sept. 23, “The Human Side of Intelligence: Resilience and Operational Capability.” Info and registration here.

📍 The Global Investigative Journalism Network is holding a free webinar on Sept. 22, “Silencing Journalists Online: How Copyright Is Being Weaponized Against Investigative Journalists.” Info and registration here.

📍 Futurism published an investigation into Brown Brothers Media. The headline sums it up nicely: “How Three Brothers Built an AI Slop Empire by Buying Legitimate News Sites and Turning Them Into Zombie Content Farms That They Say Get 50 Million Page Views per Month.” Authors Maggie Harrison Dupré and Jon Christian said the operation generated more traffic than NPR, CBS News, and Forbes.

📍 ABC News Verify found that an Australian-presenting YouTuber promoting the far-right party One Nation was actually a Brazilian man with an AI-dubbed Australian accent.

📍 Facta uncovered a churn-and-burn Google Ads scheme that impersonated Italian media with false stories that served as a lure for an advance payment scam.

📍 The Verge dug into a network of dating apps that were populated by synthetic women and paid gig workers. The article builds on Anthropic’s recent disclosure about the operation.

📍 Researcher and friend of Indicator Benjamin Shultz published a report on the prevalence of deepfake porn targeting members of parliament of EU member states. Shultz found 138 women and nine men had been targeted out of a sample of 5,872. He also found what appear to be “nudification kits” for several MPs — websites that included bios, social media account details, and genuine images for the elected officials, plus links to AI nudifier websites. (The American Sunlight Project continues to track similar stats for US elected officials and will be conducting a review for the UK Parliament.)

📍 Aos Fatos found several Instagram accounts with a cumulative 5.6 million followers recently changed their names to support André Mendonça, a Brazilian Supreme Court justice locked in a public dispute with fellow justice Alexandre de Moraes. The largest account previously posted anti-LGBTQIA material, religious, and sports content.

📍 A study in Humanities and Social Sciences Communications used a combination of survey and data donations to analyze consumption of misinformation on Hungarian Facebook. It concluded that “from the socio-demographic factors only age was consistently associated with misinformation engagement, but membership in specific interest clusters on Facebook strongly shapes misinformation engagement.”

Please upgrade to access all of our content, including our how-to guides and Academic Library, and to our live monthly workshops.