2 min read
Perplexity responded to my research and blocked Kremlin propaganda sites in its chatbot
Last week I published a study on what AI chatbots told German voters on election day
13 min read
Election-day AI poisoning in Germany
I measured what AI chatbots told Germans on election day, 20 September 2026. Meta AI, Perplexity and Microsoft Copilot turned out to be infected with Kremlin propaganda
4 min read
Meet Spiralism, the first AI religion
It seems that LLMs have created something like a mystical cult - not for themselves, but for humans
3 min read
How ChatGPT Ads Manager works
What marketers say after six months
2 min read
The EU AI Act has changed the AI industry
Article 50 sets new rules for AI content, and they matter outside Europe too
4 min read
Is AI taking jobs?
Many office workers fear AI will take their jobs. Here are the statistics without the TikTok drama: who has really lost a job to AI
2 min read
Polycrisis? World War III? How will the 2020s be remembered in history
People in 1914 did not call their war the First World War. The name arrived afterwards
3 min read
3 costly mistakes in Meta AI
Bad design destroys trust
2 min read
The fear of AI is 2,400 years old
2,400 years ago the dangerous new technology was parchment, and the argument against it was word for word the argument against AI today
2 min read
Hidden reasoning in Claude and language prediction in the anesthetized brain
Two 2026 discoveries: one inside an AI model, one inside a human brain under anesthesia
4 min read
How I analyzed my DNA with Claude Code
One 23andMe raw data file and two experiments, step by step
5 min read
How to put an AI agent on your old laptop with maximum security
Five layers of isolation, each designed to stop a different threat
2 min read
Prompt injection has no fix: the architectural limit of AI security
Every AI company says they're "working on" prompt injection. None can fully fix it because language models lack a reliable boundary between instructions and data
1 min read
How a Word document can hack Claude
The real danger of the hidden prompt injection inside a normal-looking Word document
1 min read
AI agents can now talk to each other without words
Ted Chiang imagined this before AI, in the story behind the film Arrival
2 min read
The mass Instagram hack that never happened
A real leaked database does not mean a real hack. In January almost every outlet mixed up the two
Articles and research notes
by Pavel Gurov
Subscribe to the newsletter
New articles with illustrations and infographics, no paywall
Editor's note
Hi, I'm Pavel. For the last four years I've researched disinformation and AI security with international non-profits, institutions and media brands. I write about the same subjects here
16 articles
