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 crunched the data from 19 online OSINT toolkits and produced a list of the top 50+ most recommended OSINT tools. HaveIBeenPwned, Intelligence X, and Shodan tied for the top spot, each showing up in 13 of 19 toolkits. Read the full list here.
Alexios wrote a guide to red teaming AI tools for safety. It's a combination of simple frameworks and snippets of worked-through examples drawn from his experience at Google, running a pro bono clinic at Cornell Tech, and doing this work as a journalist.
Indicator’s work was also quoted in this Wired article about Meta ads containing AI-generated child sexual abuse imagery.
One more thing: Craig is teaching a five-week webinar series with the Knight Center for Journalism in the Americas, “Search in the age of AI: How to find what (and who) you’re looking for.” You’ll learn how to use search engines, AI tools, Boolean operators, and advanced research techniques to uncover information faster, verify findings, and strengthen investigative reporting.
You'll walk away with tools and practical tips that you can apply right away, along with permanent access to the recordings and materials. Even better, the course comes with two months of full access to Indicator.
We had our first session this week but it’s not too late to sign up. For $99 you get all the webinar recordings and materials, a certificate upon completion, plus two months of everything Indicator publishes, enhanced access to OSINT Navigator, and two exclusive members-only Indicator workshops. More here.
Deception in the News
Lawfare
Explosive findings from Renée DiResta and Ronald Robertson: Grokipedia appears to have gone unedited for more than three months. Elon Musk’s AI-written, anti-Wikipedia encyclopedia is ignoring user-proposed edits and not updating high-profile pages.
Across 34,519 pages with at least one suggested edit, Grokipedia hasn’t taken any action since April, according to the research. DiResta and Robertson also looked at Wayback Machine captures of 480 pages from before the freeze period and compared them to the current text. Again, they found no changes. Even the page for SpaceX is frozen to a time preceding its blockbuster IPO.
Grokipedia
It’s unclear what’s going on. My email to xAI went unanswered. The most obvious possibility is that Musk lost interest in the project and staff was re-allocated to putting data centers on the moon, or a less petty and more profitable endeavor. (Musk has a history of making bold claims about addressing misinformation that he doesn’t follow through on.) According to Similarweb, Grokipedia drew in 6.7 million visits in June, a tiny fraction of the 3.3 billion at Wikipedia.
Perhaps Musk is content leaving a project designed to represent the world the way he thinks it should be seen as a zombie site queried only by LLM agents, imitating other influence operations. But I’d be surprised if this is curtains for Grokipedia. — Alexios
📍 In more Grok news, a federal judge denied xAI’s request to immediately pause the new Minnesota AI nudifier law that we wrote about last week. A hearing is scheduled for Aug. 19 to determine whether to provide a preliminary injunction.
📍 Disinformation about Spain’s immigration policy surged both before and after roughly 70,000 migrants entered the Spanish exclave of Ceuta from Morocco.
📍 Scammers are targeting followers of OnlyFans stars with AI-generated videos of the adult creators that promise exclusive content in exchange for a payment on Cash App.
📍 Womp womp: Google rolled back the controversial AI-powered feature that allowed users to fabricate photorealistic satellite imagery in Google Earth. The company said it would “work on implementing stronger guardrails.”
📍 AI music generator Suno announced it will adopt “new audio watermarking and fingerprinting technology” to help identify content generated with its tool. It also updated its harmful behavior policies “to make them easier to understand.” The company is facing several lawsuits.
📍 Snap said it would no longer recommend “wholly AI-generated videos” in the Spotlight feed. Besides offering the vague truism that “no detection system is perfect,” it did not specify how it will detect synthetic content. I posted a Gemini-generated video of a dancing corgi which received no label, fwiw. (I also added Snap’s stance to the Indicator guide on AI labels, which collects the relevant policies of many major platforms in one spot.) — Alexios
📍 Friend-of-Indicator Rob Leathern flagged a convincing phishing email leveraging Luma and Gmail branding.
Sponsored
Sign up for FREE weekly updates about the latest viral hoaxes & misinformation trends on LeadStories.com/newsletter.
Tools & Tips
Veridion Data launched registry-lookup.com, a free site that aggregates data from global corporate registries. It has records for over 520 million legal entities in 244 countries. You can also do 5,000 API calls a month for free.
A company post about the tool says the data includes “legal name, registry number, jurisdiction code, status, incorporation date, legal form, registered address, and identifiers like tax IDs, VAT where the registry publishes them.” It has a nice interface with autocomplete and other features.
By comparison, the free OpenCorporates site says it has data for over 220 million companies. But I’d definitely still check both. — Craig
📍 The team behind the Wayback Machine launched Ask Me Twice, “A longitudinal archive of AI chatbot responses.” It’s an early-stage experiment that helps you see how different models respond to questions, and if/how the answers change over time.
📍 My OSINT Training added two OSINT bookmarklets to its ever-growing collection of free, useful tools. One resolves GPS coordinates in Baidu Maps and the other uses coordinates to build a search grid that you can download as a KML file. If you want to learn more about how to use bookmarklets, check out our guide, “How bookmarklets can reveal hidden online profile info and speed up OSINT investigations.”
📍 Mike Reilley published a free excerpt about data scraping from the new edition of his book, “Data + Journalism: A Story-Driven Approach to Data Storytelling.”
📍 The Global Investigative Journalism Network published, “Do You Have ‘the OSINT Mindset’?: Try Our Quiz for Those Who Like ‘a Little Light Online Sleuthing’.” It’s a Google Form-based quiz that tests your reverse image search skills, among other things.
📍 The OSMOSIS community published a white paper from Yalda Paigeer titled, “AI Doesn't Replace Tradecraft—It Reveals It.”
📍 The OSINT Newsletter published, “Everything Must Go (Including Their OPSEC): OSINT on eCommerce and Marketplace.”
📍 The Reuters Institute for the Study of Journalism published a Q&A with Malachy Browne, cofounder of the New York Times’ visual investigations team and a longtime leader in digital verification. Here’s an excerpt:
Q. For journalists who want to get better at visual investigations, what are some tools or techniques they can start exploring?
A. For me, the process is very much about collecting information and mapping things out in space and time. I use tools like Google Earth to plot where things happened, key events, people, pieces of evidence, videos, and photographs, and then combine that with a chronology of events. Those two things repeatedly help me answer questions or reveal the story. Take the example of a recent strike in Iran. Videos emerged on social media within about 30 minutes of each other. Once I verified where they were taken, I could confirm that both a school and an Iranian military facility beside it had been struck in the same initial wave of attacks.
That immediately suggested certain possibilities about who was responsible, because those actors were already targeting military facilities in the area. Then, by comparing the videos with satellite imagery, you can determine that these were highly precise strikes, likely involving precision-guided munitions.
By mapping everything out and continually asking questions, you open up avenues for further investigation and eventually questions for the authorities involved. Going through that process allows you to build confidence in your conclusions.
Events & Learning
📍 The University of Southern California is hosting the second Open Source Journalism Conference. It’s Oct. 9 and 10 in Los Angeles. Early bird pricing is $100 for working journalists and $30 for student journalists. Info and registration.
📍 Craig is taking part in two events in Australia this fall. 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’s giving a talk at the Australian OSINT Symposium, "Knowing When You Know: A Framework for Evaluating Digital Indicators." Info and registration is here.
Reports & Research
📍 The UK AI Security Institute found that Anthropic’s Mythos 5 created fake accounts to try and pressure a GitHub repository maintainer into approving its code edits. The changes would have introduced malicious code. (More in the technical report.)
📍 The AI-generated MAGA commentator Marwane has more than 36,000 followers on YouTube, and received nearly half a million views on each of its two most popular videos. Riddance found “Marwane” mostly vlogs about the SAVE America Act, the Trump administration’s attempt to overhaul election law.
📍 Researcher Jonathan Corpus Ong writes that the anti-disinformation law approved in the Philippines “misunderstands a decade of disinformation research in the Philippines, as well as global case studies on digital repression.”
📍 The University of Pennsylvania’s Center on Media, Technology, and Democracy curates a valuable library of studies on LLMs and civic discourse (h/t MinJi Song).
Want more studies on digital deception? Paid subscribers get access to our Academic Library with 75 categorized and summarized studies:
One More Thing
X named Nikia Bier as head of product a little over a year ago. Since then, we’ve catalogued his efforts to try and rein in the platform’s outrage-and-misinformation-for-money incentives, improve transparency, and otherwise undo some of the detrimental changes implemented by Elon Musk after he acquired the platform.
Back in March, I wrote that it was an open question “whether Musk’s embrace of increasingly extreme politics, AI slop, and sexualized imagery will undermine or derail Bier’s product roadmap.”
Well, this week Bier announced he’s leaving his full-time role at X. It’s unclear what led to his departure; Bier said “it's now time for me to take a breather.”
Bier was on balance a positive force for change at X. He also provided a wellspring of entertainment by publicly calling out platform manipulators who complained about being demonetized, which was both hilarious and well-suited to X. — Craig
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.
