Hackers Are Already Attacking You

Summary

This video warns you about the real security risks of running AI agents and LLM services on your personal or local machines, particularly Macs. The presenter makes a case for using disposable, resettable servers instead of repurposing spare hardware like a Mac Mini for hosting AI tools like OpenClaw. The core argument is simple: when you spin up a service and expose it to the internet, you are immediately under attack. This is not hypothetical — the presenter has demonstrated this in previous videos. One of the biggest red flags highlighted is allowing your AI agent to browse the mobile web. Mobile sites can carry malicious code designed to hijack your bot or inject harmful instructions, and if your agent is compromised on a local machine, you have no clean recovery path. You cannot just reinstall the OS quickly and get back to work. A cloud or remote server, by contrast, can be wiped and restarted in minutes. The presenter specifically advises against running any mission-critical workloads on a machine that sits on your local network and runs an LLM with open internet access. The risk is not just theoretical — the attack surface is enormous. The video wraps up as part of a larger breakdown of five reasons not to use a spare Mac as your AI agent server, with the presenter being transparent that there is no financial incentive behind the recommendation — this is purely practical security advice.