Briefing: Defending against doxxing and data brokers

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

Craig dug into the features and benefits of 12 free tools for monitoring webpages and keywords. He described what each tool does, which ones are worth your time, and how to monitor a site without the owner noticing.

Alexios found that 11,000 ads for AI nudifiers have run on Meta since February, reaching a least 2 million users in the EU and UK alone. About half of the ads were connected to repeat offenders that Meta could have detected with simple text searches.

BTW did you know you can follow us on Bluesky, Instagram, LinkedIn, and YouTube?

📍 A pro-Trump X user got 3,700 likes for falsely claiming that an old French map showed Lake Ontario was originally called Lac America. The hoax was perfectly timed to capitalize on Donald Trump’s executive order that renamed Lake Ontario to Lake America. Google was among the companies that moved swiftly to rename the lake for its US users, citing a policy of deferring to governments on contested names. Though the tech giant notably made the switch faster than the US government itself.

📍 Instagram will change the label applied to synthetic profiles from “AI creator” to “AI-generated profile.” Accounts that don’t apply the label correctly “may see limits to their profile’s reach.” (That conditional is the difference between an important change and a nothingburger, so it’s a bit disappointing that the company didn’t share more details about the policy. — Alexios.)

📍 Brazil’s Superior Electoral Court (TSE) chose not to punish presidential candidate Flávio Bolsonaro for airing an AI video of his dad Jair during a party convention. Bolsonaro senior is under house arrest and the country’s election laws prohibit the use of deepfakes in electioneering, but the court determined that the event was a private one rather than a campaign stop. (h/t Henry Ajder)

The EU Commission designated ChatGPT as a Very Large Online Search Engine and Reddit as a Very Large Online Platform (or VLOSE and VLOP, if you want to impress friends in the Euro bubble). The designations trigger new obligations for the companies to assess and mitigate systemic risks on their platforms, including with respect to disinformation.

📍 A man who allegedly scammed several women out of more than $1.3 million by posing as an NFL player cited a Google AI Overview on social media that falsely corroborated his claims. “That’s Google. That’s not me, that’s Google,” he said in an Instagram story.

📍 Press Gazette keeps finding tech journalists who don’t actually exist but somehow get published in Forbes, Entrepreneur, Information Age, and The Next Web.

📍 The Electronic Frontier Foundation released a two-part guide to online safety and doxxing. Part one “focuses on prevention and ways to reduce your overall footprint.” Part two looks at “incident response, as in, steps to take if you're in the midst of being doxxed.”

The first article provides an overview of common sources of open source information, including data breaches, public records, and reverse image search. “OSINT sits at the heart of doxxing campaigns but is also an important part of the process of preventing them,” it says.

📍 On a related note, remove-your-data is a new, open-source AI agent skill that says it can “remove a person's listings from people-search sites and data brokers without paying deletion services.”

📍 Legendary OSINT author and trainer Michael Bazzell uploaded an archive of IntelTechniques blog posts dating back to 2020. You can read them here.

📍 The International Consortium of Investigative Journalists released Nuditag, an open-source tool that scans a directory for images and videos and automatically tags them as NSFW or SFW based on their content. “No GPU, no API key, nothing uploaded: every file stays on your machine,” according to the GitHub page.

📍 I recently installed the Compound Writing plugin for Claude (it also works in Codex.) It was built by Katie Parrott, a staff writer for Every who publishes interesting pieces about using AI in her work. Compound Writing is a “context-first writing toolbox for developing ideas, drafting, revising, stress-testing, and preparing work for publication without sanding off the writer.” Here’s her post about why she created it, and Parrott’s new guide to Compound Writing.— Craig

📍 Generative AI in the Newsroom announced the winners of the Agentic AI Investigative Challenge. Winners Emma Bingham and Ian Clester built a “newsroom system for lead generation. Their workflow begins by profiling and cleaning data, loading everything into a database with links back to source records. From there, it breaks discovery down into two approaches, encoded in Agent Skills.”

📍 Hackers Arise published, “What Open Maps and Trackers Reveal About Rail Networks.”

📍 Nico Dekens wrote, “When the Source Attacks Back: Prompt Injection Is Coming for OSINT.”

📍 Chris Poulter of OSINT Combine gave a keynote at the recent SANS OSINT summit, “GenAI as an OSINT Force Multiplier.” You can watch the video below. (via Brad Snyder)

📍 The University of Southern California is hosting the second Open Source Journalism Conference on Oct. 9 and 10 in Los Angeles. There’s a conference track and training track. Topics include using open sources and satellite imagery for uncovering data about global nuclear forces, and using remote sensing data to investigate a mass grave. Early bird pricing — which ends soon — is $100 for working journalists and $30 for student journalists. Info and registration.

📍 The EU DisinfoLab is hosting a free webinar on Sept. 17 with Factiverse and SimPPL, “Reject Then Launder: Catching AI Models Validating Extremist Narratives in Crisis.” Registration and info is here.

📍 Craig is speaking at two events in Australia next week. On Sept. 8 he’s giving the A N Smith Lecture in Journalism at the University of Melbourne on the topic of "Industrialised deception: the media’s truth crisis." You can register to attend for free. On Sept. 10, he delivers a talk at the Australian OSINT Symposium, "Knowing When You Know: A Framework for Evaluating Digital Indicators." Info and registration is here.

📍 Full Fact tracked AI chatbot responses to disputed images, logging 39 major errors across five months of testing. More interesting to me in terms of the (unpaid) value fact checks might provide is the fact that some responses improved after the fact check was published and could be cited. — Alexios.

📍 Related: NPR and NewsGuard found that major AI chatbots “mostly pushed back against false narratives spread by foreign states,” performing better than the AI overviews at the top of Search results for the same queries.

📍 CNA News analyzed Zoom call footage where scammers used edited video and synthetic audio of Singapore’s prime minister to pull off a S$4.9 million fraud.

📍 Semafor reported on a startup that bypasses AI-detection filters by pairing AI-written, anti-Democrat scripts with real human actors posing as political news presenters. “Over the last several months, a new generation of YouTube growth-hackers have flooded American politics with sensationalized, low-quality content, reaching millions of viewers with nearly identical videos taking aim at progressive politicians,” wrote Max Tani.

📍 A new preprint found that X users who have a Community Note added to their tweet see a reduction in future posting, including of misleading content. But that was only among occasional posters. Repeat offenders on the other hand “ exhibit no comparable content improvement, and instead publish more highly misleading posts, cite lower-quality domains, and post more political content.”

📍 Last week, X announced it disrupted a Chinese online influence operation pushing anti-data center narratives, earning praise from the US State Department official in charge of public diplomacy. Renée DiResta and Joe Bodnar published smart takes on the takedown. DiResta called out the hypocrisy of pro-Trump figures endorsing research into influence operations after they had demonized such work. She also noted that X provided a fraction of the transparency it offered in prior reports that were delivered in partnership with the Stanford Internet Observatory, where DiResta used to work. Bodnar wrote that the renewed interest “in foreign disinfo is being driven by overhyped reports about Chinese interference in data center debates,” noting that the same operation was characterized as a low-level campaign in a more detailed OpenAI report.

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