Is a secure AI assistant possible?
This viral AI tool is handing over your hard drive to LLMs...
Deep Dive
Independent developer Peter Steinberger's OpenClaw project has gone viral, allowing users to create 24/7 AI assistants with access to emails, hard drives, and payment info. Security experts are alarmed, with China issuing a public warning about vulnerabilities. The tool lets any LLM act as a persistent 'mecha suit' with memory and task automation, but risks include data extraction, hacking, and catastrophic errors like wiping entire drives. Steinberger warns non-technical users to avoid it.
Why It Matters
The explosive demand for powerful AI agents forces a major security reckoning for the entire personal assistant industry.