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 27 tips for investigating mobile apps. They describe how to use free tools and techniques to uncover app ownership, analyze store metadata, spot red flags, and more.
Alexios interviewed Roblox vice president of safety policy Eliza Jacobs. They discussed the challenges off-platform behavior poses to on-platform moderation, how AI complicates age verification, and whether strictly defined disinformation is a major risk on Roblox. Also: jigsaws, and the awesomeness of Daphne Keller.
In other Indicator news: Alexandria Ocasio-Cortez cited the main findings from our 2025 report on the business model sustaining AI nudifiers during an Instagram story promoting the DEFIANCE Act. You can find all of our nudifier reporting, including the (just updated) map of reported cases of deepfake nudes targeting school communities, at this link.
Deception in the News

📍 LinkedIn chief product officer Hari Srinivasan posted that the platform is testing new classifiers to reduce the visibility of “AI slop” comments. Srinivasan called this “an area that pretty clearly needed attention.”
📍 Meta’s AI agent Muse has “burrowed into encrypted personal messages, supplied made-up information about purchases and sent users chats seemingly meant for other people,” according to internal company posts reviewed by Reuters.
📍 Aos Fatos reports that far-right legislators and influencers in Brazil falsely claimed that the country’s Digital Markets Bill would enable censorship of online content.
📍 The AI podcast company Inception Point AI — whose sloppy output Alexios binged last year — has rebranded its business as Liquid Studios and pivoted to turning other companies’ podcasts and newsletters into videos. RIP, “How to Cheek Kiss.”
📍 The UN-backed Priorities for Action on AI for Climate Information Integrity were launched at the COP31 climate change summit in Türkiye last Friday. They set out priorities for AI actors, such as relying on authoritative scientific sources and not rewarding “agreement above accuracy.”
📍 This week’s brand new sentence is courtesy of 404 Media: “a man in South Carolina named Joe Thomas has used Meta’s Muse agent to sue Meta after the company’s content moderation filters locked his epoxy countertop business’s Facebook page after he gave his agent, ‘Baby Mia,’ access to the account.”
📍 A company linked to an online scam network we investigated pleaded guilty this week to trafficking synthetic opioids. eWorldTrade, an online B2B marketplace, admitted to conspiring to distribute synthetic opioids for unlawful importation into the US. It agreed to pay a $1.35 million fine.
In May 2025, we reported that eWorldTrade is connected to a group of people and companies in Pakistan that used fake reviews on Trustpilot and Meta's platforms to promote its scammy ghostwriting and publishing services. Customers alleged in lawsuits that they didn't receive what they paid for. The DOJ's announcement does not mention the network's other businesses. — Craig

Sponsored
Vanity Fair’s New Show, Where Nothing Is Off the Table

On Fair Game from Vanity Fair, join Mark Guiducci for in-depth and far-reaching conversations with the most interesting characters from the worlds of Hollywood, Politics, Style, Silicon Valley, and beyond.
From acclaimed actors and political figures to tech leaders and style visionaries, Fair Game goes beyond the headlines to uncover the people, personalities, and ideas driving the cultural conversation.

Tools & Tips
Milo von Klis and Ben Polonsky are second-year cybersecurity students at Champlain College. They won the Most Valuable OSINT award at a recent TraceLabs Search Party CTF and published a helpful post about what they did.
It details how they used Claude to speed up data broker lookups and court record searches, among other tasks, to assist with a missing person investigation. From the post:
After discovering the legal name from my large repository of images and links, I used my Claude skillset to connect to the local municipality's and state court records and pull the relevant details from each of them: defendant, crime, outcome, etc. The non-automated way of doing this was such a pain that automating it with AI saved us so much time. After putting this into a timeline, we uncovered details of the target's life before their disappearance.
Their submission included this timeline mapping court records alongside social media activity:

A timeline from the team’s winning submission mapping court records and social media activity
“Sometimes I think we’re doing cyberstalking that really isn’t going to help anyone, but I have to remind myself that any information we find could genuinely help find these people,” Polonsky wrote. — Craig
📍 Mutuals is a new tool that takes two public social media accounts and identifies the accounts that connect them, such as shared followers and other network connections. You can also check out Plessas Labs’ Friend Overlap tool.
📍 Interchange is a free tool for tracing and visualizing cryptocurrency funds across multiple blockchains.
📍 The Company Relationship Explorer from Lazlo Intelligence maps out a company's corporate connections. (via Mario Santella)
📍 Dennis Keefe built Bluesky Handle Checker, a username enumeration tool. “Give the tool a list of possible handles and see which ones return a public Bluesky profile,” he said. “You can then spend your time reviewing the results that matter.”
📍 ChatCrawler is “a catalog of [Telegram] channels and groups, rankings and statistics of every venue.” You can use it to track historical usernames and check when Telegram channels and groups may have changed their names. (via @techniquesosint)
📍 Cyber Detective shared a curated list of image search tools from their massive GitHub OSINT repo.
📍 Public Insights is a UK public records search platform that offers five standard searches per month. The service says it has more than 52 million records across multiple datasets, including planning applications, court records, insolvencies, probate records, and more. Info here.
📍 OSINT-UI published “Bluesky OSINT: How to Investigate an Account, Handle History, Blocks and Domain.”
📍 UserSearch published “Google Dorks for OSINT: A Quick Reference for Research Teams.” It covers which Google operators still work, how Bing, DuckDuckGo, and Brave compare, and more.
📍 Mike Caulfield shared his “model library” approach to prompting as a way to verify potentially misleading or out of context images.
📍 The Data and Research Centre published “Inside Deutsche Bank: Making 15 million files searchable, and helping journalists find their stories.”
📍 Ryan McNeill, deputy data editor at Reuters, gave a free workshop for the Global Investigative Journalism Network on “Using Satellite Imagery to Uncover War Crimes and Environmental Damage.” Watch it:

Events & Learning
📍 Next Sight is hosting a free webinar on Dec. 9, “Agentic OSINT: free online training for investigators.” Info and registration here.

Reports & Research

📍 Vietnam-based Facebook content farms are pivoting to content about the Spanish election, according to a new analysis by Maldita.
📍 The nonprofit Nexus Project says Meta’s move to a more hands-off moderation approach has made Instagram feeds more prone to antisemitism. In a new report, it describes setting up a new Instagram account, searching for the word “synagogue,” and watching a few videos. Instagram then recommended 15 videos, 14 of which were antisemitic.
📍 A UNESCO and University College London report tested up to 33 AI models on climate change questions and found that they performed well on settled science, but did no better than random guessing when asked how two climate policies would work in combination. In a sycophancy test of 15 models, every one eventually abandoned a correct answer when researchers pushed back with climate contrarian arguments, and some went on to endorse false claims.


One More Thing
Menswear designer Reese Cooper noticed something was off about the way his clothes looked on Ssense, a high-end fashion website. In a series of Instagram stories, he accused the retailer of using AI-generated imagery to display his products.
Cooper shared the left-hand image below to show what one of his jackets looked like on the Ssense site. The right-hand image is what it actually looks like.

“Seeing the things you've made photographed by SSense used to be such an incredible feeling,” he wrote on one image. “This is just insanely disappointing.”
Earlier this year, The Logic published an article headlined “Ssense has laid off photo and make-up teams and says AI will do much of their work.”
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.




