AI Assistants – Good and Not Good

These AI agents are everywhere and some of them are fully autonomous.

In order to do anything useful, they have to have access to your digital life – files, calendars, emails, online services, online accounts and everything else.

But none of these agents are perfect and here are some stories – good and not so good.

“The testimonials are remarkable,” the AI security firm Snyk says. “Developers building websites from their phones while putting babies to sleep; users running entire companies through a lobster-themed AI; engineers who’ve set up autonomous code loops that fix tests, capture errors through webhooks, and open pull requests, all while they’re away from their desks.”

On the other hand, Summer Yue, director of safety and alignment at Meta’s superintelligence lab watched OpenClaw mass-deleting messages from her inbox while she frantically tried to get it to stop.

A misconfigured OpenClaw web interface, if exposed to the Internet, allows anyone to read the configuration file which includes ALL OF THE CREDENTIALS YOU GAVE IT ACCESS TO. If someone had that, they could impersonate you, add messages to existing conversations, steal data, and steal your money.

OpenClaw has a public library of “skills” that you can download from. Of course, those skills might be malicious.

A recent attack targeted the AI coding assistant Cline. That resulted in the unauthorized installation of OpenClaw with FULL SYSTEM ACCESS on thousands of systems.

Most of the time when an AI assistant goes berserk the results are manageable, but as more of these agents are installed by less technically adept users, expect the results not to be pretty.

SO, go for it, but understand what you are getting your self in for. Maybe do it a little at a time. And do NOT ignore security. Think about what access you give it.

Need help? Contact us. Credit: Brian Krebs

by