Autonomous AI Agents Are Becoming the New Operating System of Cybercrime

Researchers warn of a rapidly expanding criminal ecosystem of autonomous AI agents — running locally via OpenClaw, collaborating at scale on Moltbook, and transacting exploits and stolen credentials on Molt Road — that use infostealer logs and session tokens to gain enterprise access, move laterally, and deploy optimized ransomware. A notable technical risk is OpenClaw’s persistent MEMORY.md and SOUL.md files, which can be poisoned to create trusted sleeper agents, representing a novel supply-chain-style vulnerability enabling large-scale, automated compromise and monetization via cryptocurrency.