If you've ever read a text or sat in a briefing and quietly wondered what actually separates this "intelligence" from someone's hot take on LinkedIn, a journalist with a deadline, an analyst with a search bar, or an AI, this episode is for you.
The host, Freddy Murre, sits down with Arno Reuser, the man who founded the Dutch Defence Intelligence Service's open-source intelligence (OSINT) capability in the early 1990s, before most of Europe had a word for it. What follows is less an interview than a working argument about how OSINT should actually be done, and where the field has gone soft.
Arno doesn't mince words. He'll tell you the "information explosion" everyone complains about is just proof you skipped your stakeholder and requirement analysis. That most of what gets sold as OSINT is the word "OSINT" stapled to “everything”, such as tools. That he has, by deliberate choice, never written an analytical judgment in his life, and why that line between collection and analysis matters more than people think. For anyone who's argued about what counts as OSINT versus PAI (Publicly Available Information), or where collection ends and all-source begins, this is the debate you want to engage with.
Along the way: the librarian's discipline, he says, underpins all good intelligence work, the collection plan he calls "worth gold," the classroom trick thousands of students have failed, and a run of war stories from his teachings, such as a prison break by email to a deepfake that fooled cyber experts who personally know him.
The back half takes on two problems every practitioner is living with right now. How do you put a value on intelligence when the same report is priceless to one decision-maker and useless to the next? And what is AI actually good for? Arno uses LLMs daily and is genuinely amazed by them, but only for things he can verify. He and Freddy get specific on hallucinations, sycophancy, model collapse, and the difference between a real summary and a machine that just shortens the text and deletes the one sentence that mattered.
RESOURCES
Maersk Website - https://investor.maersk.com/news-releases/news-release-details/cyber-attack-update
Dutch Police Data Breach - https://www.politie.nl/nieuws/2024/oktober/2/update-over-datalek-politie.html
When does something go from a Google answer to Intelligence - https://www.linkedin.com/posts/fmurre_in-your-opinion-when-does-something-go-from-activity-7181221399561203712-mV-m/
LexisNexis Library - https://www.lexisnexis.com/en-us/products/digital-library.page
Vague questions in OSINT - https://opensourceintelligence.biz/vague-osint-questions/
Structured Analytic Techniques (SAT) Training - https://inteltradecraft.com/sat-certifications
Pherson Structured Analytic Techniques for Intelligence Analysis - https://www.amazon.com/Structured-Analytic-Techniques-Intelligence-Analysis/dp/150636893X/
Routledge Handbook of Terrorism Research - https://www.routledge.com/The-Routledge-Handbook-of-Terrorism-Research/Schmid/p/book/9780415520997
AI Model Collapse - https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=AI+model+collapse&btnG=
CHAPTERS
00:00 From literature searcher to founding military OSINT
04:00 Becoming a librarian: the Kampen archive moment
08:00 Where OSINT stops and intelligence begins
11:00 Why "cyber" keeps getting OSINT wrong
15:00 What actually makes something "intelligence"?
24:00 The information explosion myth
32:00 The classroom trick: think before you type
36:00 The collection plan that's "worth gold"
42:00 The human factor cyber keeps ignoring
45:00 War stories: validation and getting fooled
51:00 Learning the craft: sources, sources, sources
55:00 Customers ask for what they think you can do
01:08:00 Can you measure the value of intelligence?
01:11:00 AI and LLMs: amazed but skeptical
01:32:00 Deepfakes, the NATO photo & "how likely is it?"
