Using GrokBot as an OSINT Tool to Trace the Origin of Forged AfD Emails
FlorianGallwitz · x · 2026-10-23
A concrete demo of GrokBot performing a fact-checking OSINT task: tracing where forged AfD emails first appeared. Per the investigation screenshots, the fakes first surfaced on Oct 20 at 16:36 Berlin time via @RechercheRecTv — an account created Feb 2026, based in Switzerland, with 15 followers that had posted genuine leak-index screenshots that morning. The spread chain: reposted at 17:04, hit 559 likes on Bluesky at 18:11, then went wide on X from 19:07. The author manually verified 2,750 images via OCR and image comparison, confirming no earlier matches, and traced a second variant with a Hitler portrait to a deleted Green politician post. Also notable: handing inaccessible content to a second agent for read-only review.
More from coding & agent
- Agent Prompts Have No Guarantees: Injected Text Can Override Your Instructions — mmitchell_ai · 2026-10-23
- Grok Bot adds 1Password integration so the agent can sign in for you — belce_dogru · 2026-10-23
- Notable AI safety lead warns: agents take instructions from web pages and emails, not just you — mmitchell_ai · 2026-10-23
- Grok Bot is more than a chatbot: 8 tasks you can hand off to xAI's always-on agent — nima_owji · 2026-10-23
- AI agents fuel real-world intrusions: 2,250 M365 tenants hit, 100+ firms compromised — cyb3rops · 2026-10-23
- New Context Engineering Handbook chapter offers a systematic guide to agent instructions — dbreunig · 2026-10-23