Contexte

These sources describe OpenClaw, an open-source autonomous AI assistant designed to execute complex workflows across messaging platforms like WhatsApp, Slack, and Discord. Originally created by Peter Steinberger, the project achieved massive viral popularity in early 2026, leading to its transition toward an open-source foundation under OpenAI. Technically, the system is distinguished by its transparent memory architecture, which utilizes human-editable Markdown files and vector databases to store long-term information. While praised for its automation capabilities in fields like ESG data collection, researchers have raised serious alarms regarding critical security vulnerabilities. Specifically, the “ClawHub” marketplace was found to host hundreds of malicious skills capable of delivering malware and exfiltrating sensitive data. Users are also cautioned about extreme token consumption costs and the risks associated with granting the AI deep system permissions.

Chapitres

  • 0:00 — Introduction OpenClaw
  • 1:08 — Danger du pouvoir total
  • 2:15 — Chaîne d’approvisionnement empoisonnée
  • 3:20 — Agent retourné contre utilisateur
  • 4:40 — Solutions de protection

Sources

Voir les 14 sources restantes