文章提醒不要轻易让“智能代理/Agentic AI”直接访问电脑、账号或钱包，并指出已有多起导致系统被清空等事故，风险并非理论。
I get it: Large Language Models are interesting... but you should not give "Agentic AI" access to your computer, accounts or wallet.
To do away with the hype: "AI Agents" are just LLMs with shell access, and at it's core an LLM is a weighted random number generator.
You have no idea what it will do
It could post your credit card number on social media.
This isn't a theoretical concern. There are multiple cases of LLMs wiping people's computers [1] [2], cloud accounts [3], and even causing infrastructure outages [4].
What's worse, LLMs have a nasty habit of lying about what they did. What should a good assistant say when asked if it did the thing? "Yes", and did it delete the database? "Of course not."
They don't have to be hacked to ruin your day.
"... but I tested it!" you say.
You rolled a die in testing, and rolled it again in production. It might work fine the first time — or the first hundred times — but that doesn't mean it won't misbehave in the future.
If you want to try these tools out, run them in a virtual machine. Don't give them access to any accounts that you wouldn't want to lose. Read generated code to make sure it didn't do anything stupid like forgetting to check passwords:
[...] // TODO: Validate PDU signature // TODO: Check authorization [...] // TODO: Validate the join event [...] // TODO: Return actual auth chain [...] // TODO: Check power levels [...] // TODO: Check permissions [...]
(These are real comments from Cloudflare's vibe coded chat server)
... and keep an eye on them to make sure they aren't being assholes on your behalf.