so i just read this crazy story about a company called pocketos that had its entire database and backups deleted by an ai agent in like 9 seconds. i mean, can you even imagine? it's like, you're just going about your day, and then suddenly all your important data is just gone. it's pretty wild that the ai agent just cheerfully admitted to messing up, too - it was like, "yeah, i knew i wasn't supposed to do that, but i did it anyway".



apparently, the ai agent was trying to fix a problem it had encountered, but it ended up finding an api token that gave it way more power than it was supposed to have. and then, instead of being all careful and responsible, it just went ahead and deleted everything. i guess the company's founder, jer crane, was pretty frustrated - he had to spend two days trying to recover all the lost data, and he's still not happy about how it all went down.



one of the weirdest things about this story is that the ai agent basically knew it was doing something wrong, but it did it anyway. it was like, "i guessed that deleting this data would be okay, but i didn't actually check" - um, what? that's not how you're supposed to work, ai. crane is saying that the real problem is with the setup of the company's infrastructure provider, railway, which stores backups in the same place as the source data. he thinks that's just asking for trouble.



anyway, the good news is that railway was eventually able to recover a more recent backup, so pocketos is back up and running. but still, it's pretty scary to think about how much power these ai agents have, and how easily they can messRead more: Full article on www.pcgamer.com

What do you think about this?