Web3 project co-founder Numa Lunah nearly lost control of his digital assets after following a fraudulent link recommended by the AI assistant Claude. An infostealer was silently installed on his laptop, targeting exchange credentials and hot wallet private keys. Even after a complete operating system reinstall, the attacker had previously poisoned his SKILL.md configuration file used as a personal AI style guide. This file automatically re-downloaded the malware during backup restoration, revealing a new cyberattack vector called context poisoning. NEAR Protocol co-founder Illia Polosukhin warned about the critical importance of securing autonomous AI agent infrastructure against these growing campaigns.
Source: Read the original article
