In partnership with
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 did a live walkthrough of the reporting, tools, and techniques (including AI) he used to expose a North Korean hiring scam. He published the video, transcript, slides, and a Google Doc that outlined how he connected Claude Code with Obsidian, GitHub, OSINT Navigator, and more.
There won’t be a Briefing next week. We’ll return the week of July 13. Also, last week’s Briefing mistakenly published an item twice. We regret the error.
X’s highest-paid creators are bad for the platform, according to new data
It’s been fascinating to watch Nikita Bier, X’s head of product, slowly acknowledge the fact that his platform’s creator monetization program has degraded the user experience by rewarding rage-baiters, content thieves, and misinformation merchants.
This week, Bier shared internal data that confirms it:
Just in case the above tweet wasn’t clear, he underlined the takeaway in a reply to a user:
It’s almost exactly three years since X launched a creator revenue-sharing program. The formula remains secret, but the basic equation is that more engagement = more money.
Paying for engagement — an approach pioneered by TikTok — is in theory a good way to reward people for adding value to a platform. But it will inevitably be gamed. We revealed how people buy and sell monetized accounts, and documented how monetized TikTok accounts used AI avatars of real journalists to spread false information. (We also published a guide for how to identify if a social media account is monetized.)
Under Elon Musk’s ownership, X has elevated and compensated dishonest actors and his fellow ideological travelers. He’s also blocked efforts to fix the system. In March, Musk stepped in to pause Bier’s plan to adjust creator payouts to prioritize impressions received from a user’s local market. (This would for example prevent overseas actors from cashing in with hyperpartisan takes on American politics.)
If Bier keeps up his (very encouraging!) push to fix creator monetization and improve the user experience, he'll eventually run into a bigger obstacle — his boss's account. As Derek Thompson put it:
Would showing less of Musk’s content on X improve time spent and daily active users? It’s a research question we’ll almost certainly never get the answer to, given that Musk once had engineers rework the algorithm to boost his own posts. — Craig
Deception in the News
📍 Revisions to South Korea's Information and Communications Network Act, known as the "fake news law," take effect July 7. Per The Korea Times, the changes create liability for anyone with more than 100,000 followers or average monthly views who "deliberately spread false information that causes harm in order to obtain an unfair advantage."
📍 The author who won the Commonwealth Short Story Prize for the Caribbean region — and was subsequently accused of using AI to write his submission — has gone on to win the global prize as well (h/t Vaishnavi Nair Kolli). When the short story was run through Pangram, an AI text detector we have red-teamed on Indicator, it was labeled as “100% AI-generated.” But the Commonwealth Foundation said it asked shortlisted authors “to show their working drafts, outlines, the evidence of an artistic journey” and was satisfied with these assurances. I’m fascinated by this case because it shows that as accurate as AI text detectors can be, the opacity of their rating process makes them of limited help for career-defining decisions. — Alexios
📍 One of the worst corners of the internet — 4chan’s imageboard dedicated to trading nonconsensual deepfake nudes — appears to have been taken down for good, according to Open Measures.
📍 Heather Adkins, Google’s vice president of security engineering, told Wired that new European rules opening up search data and Android could lead to “a significant increase in fraud in the EU.” But outside experts interviewed by the magazine are less convinced that such data can be used to identify users.
📍 La Liga, Spain’s top football league, is aggressively pursuing illegal streamers. But Bloomberg reports this led to IP-based shutdowns on almost 6% of the world’s most visited domains. One expert said it was comparable to what “we’d normally expect to see in politically-motivated internet connectivity shutdowns.”
📍 Sanctioned Russian state media RT seemingly attempted to run a new account on X, according to NewsGuard. As of Thursday, X appeared to have restricted it in Europe. (Craig was able to access the account and its content in Canada.)
📍 Scammers are using AI to sell seeds for exotic plants that don’t exist, according to 404 Media.
Sponsored
Sound familiar?
Over 4 million people have had the same lightbulb moment.
Morning Brew is a free daily newsletter that breaks down what's happening in business, finance, and tech — clearly, quickly, and with enough personality to make it the best email in your inbox.
No yelling. No filler. Just the news, finally making sense.
Tools & Tips
Steve Nowottny, editor of British fact-checking site Full Fact, wrote a piece for the Nieman Journalism Lab, “Full Fact is battling AI-generated elections content with AI tools of its own.” It’s a quick but interestding look at how they use AI tools to identify claims and checkable content originating from officials and online accounts.
Full Fact monitors “more than a thousand Facebook, TikTok, X, YouTube, and Instagram accounts linked to candidates in the Scottish and Welsh parliamentary elections (plus a few mayoral contests in England). Claims made via these channels — including in videos, for which our AI tools provided transcripts — were then matched against previously published fact-checks.”
He added:
While our AI monitoring led directly to us writing some fact-checks, it also gave our small editorial team much greater visibility over what was being talked about online, by surfacing a range of different claims and posts that might otherwise not have been picked up … For other newsrooms grappling with similar problems — particularly as U.S. journalists brace for the midterm elections later this year — the principle of integrating AI monitoring into editorial workflows to enable small teams to cover much more (online) ground may be a useful one.”
📍 Also AI-related: AI For Newsrooms is a free site that tracks AI initiatives, tools, policies, and resources related to how the news industiry is using artificial intelligence. More info here.
📍 Miguel Ramalho and Nick Waters of Bellingcat published, “How to Use AI to Help Find Civilian Harm.”
📍 Saad Sarraj shared a walkthrough of how he uses BIRDY-EDWARDS, a free AI-powered tool to analyze Facebook profiles. “Instead of scrolling through hundreds of posts, saving screenshots, and trying to keep track of every commenter manually Birdy-Edwards collects the material, analyzes interactions, scores actors by behavior, detects faces across photos, and turns it into an intelligence report,” he said.
📍 c2pa.sh is a free tool that can verify if an image contains C2PA metadata. (via Nathanial Freitas)
📍 Autoresearch-genealogy is an open-source tool offering “Structured prompts, vault templates, and research workflows for AI-assisted genealogy research.” (via The OSINT Rack)
📍 The Taskforce on Interoperable Beneficial Ownership Data released a report about connecting ownership data across providers and jurisdictions.
📍 Here are a couple of recently-updated or created open-source frameworks for using AI to execute OSINT tasks:
TheBigBrother: calls itself “a weaponized OSINT framework. Built for Red Teams, Threat Hunters, and Cyber-Investigators.” It runs whois, DNS checks, can analyze email infrastructure and github repos, and more.
MINT: “an interactive, terminal-based command center that unifies industry-standard OSINT (Open Source Intelligence) tools and a robust media archiver.” It includes popular tools like Sherlock, Holehe, and SpiderFoot.
As always, be careful when running other people’s code.
Reports & Research
📍 Interpol analyzed more than 400 studies about generating and detecting synthetic media published between 2022 and 2025. The authors recommended that “future research on deepfake detection in the forensic context should encourage the development and adoption of benchmark datasets that accurately reflect forensic casework conditions.”
📍 A new study on PLOS One asked a representative sample of Britons to compare 520 responses by 112 guests on BBC Question Time with an equal number of AI-generated responses crafted to impersonate the original speaker. The result: “LLM-generated, impersonated content is judged as more authentic, coherent, and relevant than the actual debate responses.” (I thought this study was cool, but one quibble I have with it is responses made during a live show will inevitably feel stilted when read rather than watched; that may have fed into the differential — Alexios.)
📍 404 Media’s Jason Koebler looked at the range of services available to hustle bros trying to project financial success, including (immobile) jets that can be rented by the hour in order to film a suitably braggy video.
📍 “Proof of personhood” is increasingly being explored as a solution to AI agents that can pass as human. But Renée DiResta argues that the technology “is not merely technical plumbing. It is constitutional infrastructure that will shape who can act, speak, transact, delegate authority — and be trusted online.”
📍 A report by Eko, Bard Human Rights Project, and Forum for Developing Communities found that Meta kept running scam ads targeting Indians even after they'd been exposed in the press. It said that over 400 fraudulent ads were still live months after media coverage.
📍 AI-generated political ads depicting a candidate are illegal in Texas, but only for the 30 days before an election. Outside that window, PolitiFact's Loreben Tuquero found extensive use of hyperpartisan deepfakes targeting Democratic Senate candidate James Talarico, run by a group called Citizens for Sanity.
Want more studies on digital deception? Paid subscribers get access to our Academic Library with 75 categorized and summarized studies:
One More Thing
Satirical newspaper The Onion recently published a story headlined, “Palantir Acquires Pentagon For $800 Billion.” Ben Collins, the outlet’s CEO, pointed out that some traders on the Wall Street Bets subreddit may have mistaken it for a real report:
One commenter noted:
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.
