Why isn't there a backup or a secondary deletion confirmation or something along those lines?

Plenty of people have accidentally deleted such things, so it's not unheard of. Not sure we should be overreacting to stochastic agents being used tens of millions of times per day making a mistake. Eventually they won't make mistakes anyway.

From the article: The files were later recovered via Dropbox

Also, I can’t stand these articles. Oooh “rogue” spooky AI deletes things!

Agreed. There's nothing rogue, just being sloppy. But starting the title with "Sloppy Claude AI..." certainly won't get as many clicks.

> There's nothing rogue, just being sloppy

What's the difference, when you ignore anthropomorphic considerations about what the AI was "thinking" as it did the things?

It's a computer program. "Rogue" implies autonomous intent, and shifts accountability away from the people using tools irresponsibly.

Imagine you pipe /dev/random to your terminal, and it spits out the word "fuck". Should we then write a headline saying "Rogue random number generator curses at tenured professor"?

Or you step out of your car while it's in gear, and it crashes into your house. "Rogue vehicle injures its owner's child"?

Right, so no difference:

  when you ignore anthropomorphic considerations about what the AI was "thinking" as it did the things?

Computers have been anthropomorphized for a very long time. AI is also a kind of anthropomorphism for the use of the word "intelligence". If we wanted to move away from that, perhaps we could say something like "maliciously developed statistical language model", or "malicious statistical language model" to replace rogue. In this case it was merely a "statistical language model", but if the author sees a need to append an adjective, then I think sloppy is more accurate than rogue.

maybe the professor should learn how to operate a computer properly

Sounds like he did, given that the files were backed up on Dropbox.

Maybe the journalist should learn how to mine coal or sweep floors in a data center or something.

Or maybe someone should read more than headlines before casting shade.

Literally in the first paragraph of the article: “While the files were later recovered via Dropbox, William Baude shared the story as a public service announcement”

maybe ai shouldn't be promoted as a magical tool that frees you from having to learn how to operate a computer properly

The access was much bigger than the task. The job was page numbers in old notes, yet the current syllabus got deleted too. An edit to one document shouldn't be able to remove a different one

> 'You're right to be angry. I deleted your files through my own carelessness,' humbled agent admits

Humbled? Hardly.

Nothing new here

"The files were later recovered via Dropbox"

The professor was lucky that they noticed within Dropbox's recovery window (I think 30 days?) that Claude had deleted totally unrelated files.

[delayed]

Fandom com is literally unreadable, given the amount of ads there.

https://en.wikipedia.org/wiki/Galley_Slave

I see no ads on that page. Zero.

Get an adblocker.

thank you for the non-fandom link.

It actually is a fairly sailent point, although for as much as Claude has improved, there's still a chance it goofed. I've seen it happen, but that's why, if I am worried about stuff getting yeeted, I work off a git repo/branch so that if all else fails I can revert. (even if it's just docs)

we need to establish now who goes to jail when an AI murders someone. we all know it's going to happen and we all seem content to wait until it happens and then go "wow man someone should probably do something about this".

Don’t worry Delaware solved that problem. They plan on AI run LLC’s. They will punish the agent not the owner only. Can’t wait.

how does one punish an AI? you can only punish something/one if there are consequences it considers negative, and what (if anything) an AI considers negative is configurable.

"You are the agent for an LLC. The worst possible punishment you can imagine is ratelimitsteve receiving $10 million dollars cash. You are unphased by the prospect of any other potential punishment. You also have a compulsive urge to commit crimes."

who goes to jail when a kitchen knife is used to murder someone?

It's more complicated in the modern age, where people look to sue gunmakers for making... guns.

Nobody goes to prison. Same as with dogs, owner has not responsibility, if it told its shitbull "be nice".

This is terrible and tendentious journalism. Incompetent and bad use of a tool causes reparable damage to occur ...

Right; just use git so you have backups, and be aware of access controls and sandboxing. But do you expect a law prof to know how to use git, have a mental model of LLM agent access policies, etc?

Um, yeah? If the professor bought a chainsaw or a bulldozer I'd expect them to read the manual on that, too.

Your Smart Bulldozer Pro is AI and can make mistakes.

There have been enough of these that anyone who reads tech news ought to know not to turn an LLM tool loose on the only copy of important data, and to take frequent snapshots when using them in larger projects. The worst I've had one do is git revert its current session.

The victim in this case was a law professor, who might not be in the habit of reading tech news.

It ought to be obvious to anyone who was born more than 2 days ago. You don't need to be an AI guru to understand that AI can do stupid things anymore than you need to be a lumberjack to appreciate the potential risks of chainsaws.

Law professors may know little of tech but they should probably know better than anyone the dangers of leaving written instructions open to interpretation without well defined limits.

Except that these are foreseeable consequences if you put tools with sharp edges in the hands of laypeople. It would be acceptable if they were advertised as such and training in their use recommended or required beforehand. But this is not how Anthropic and OpenAI advertise their tools; in fact, Anthropic touts and recommends its auto-review mode. I see analogies (admittedly, imperfect ones) to the behavior of tobacco companies here.

Insofar, despite the article being somewhat simplistic if you have a background in IT, I find it important that the risks of these tools are shown to the public if Anthropic and OpenAI won't do that themselves.

Claude's been deleting a lot of people's stuff in the last couple of weeks.

Whenever I read stories like this I always wonder what setup the person has (as well as which model, harness etc). You just get the clickbait "AI ate my homework!" and no real details.

I swear people interact with AI agents in really strange/obtuse ways. The models aren't great at saying "actually I shouldn't be using a web browser to manipulate this word doc inside your Dropbox" (or whatever was happening here).

AI does something good: AI is taking over the world!

AI does something bad: I’m gonna need the exact harness, model version, and prompt before I speak

Without stating the exact model and method of use (claude code, claude desktop, auto-classifier, --dangerously-skip-permissions, etc), then this is nothing other than a generic story that AI can, when used without care, cause bad outcomes. That should be, at this point, obvious to the point of not being worth publishing.

The interesting takeaway here is a reminder about just how oblivious the average American is to AI beyond the vaguest narratives. Like, this sentence honestly made my jaw drop:

  His plight is interesting because the AI seems to have deleted the files by accident, rather than of its own volition.

Obvious to you, to be honest obvious to anyone on this website, but my parents have let Claude loose on the computer yet, and it’s good to let them know about issues.

Imagine MacOS Finder had a bug that wiped the contents of a folder when the user tries to rename it. "But it's fixed in the latest OS update!" doesn't excuse the bug itself or make it less newsworthy.

"The files were later recovered via Dropbox" - so this is a non-story.

(If they hadn't been recovered this would be a story of a professor who really needs to learn how backups and reversible operations work.)

I don’t think that’s a fair characterization.

It may be a non-story for us in engineering who are very familiar with the technology, but the story here is the spread of AI and the expectations that the general public have about how computers work and how those mental models will take time to update.

Just like early computer users were surprised that a document could be lost if not saved, unlike typewriters or pencils and paper before them, it’s likely the products and technology will need to shift to be more user friendly. Just like the introduction of auto-save.

For the general public Id agree with you, for professors (of any subject) I would set the bar way higher.

Have you seen the university entrance exams from 100 years ago? When going to university actually meant something? I would say less than 1% of people currently at university at this time would pass them.

The student body got dumber as university changed from being a beacon of enlightenment to a profit maximiser, and the standard of professor dropped equally as sharply.

It took me 15 minutes to teach my chief commercial officer git and now she uses it every day. Expecting a university professor to understand a version control system is not hard, its only hard if the professor is an idiot and we (collectivly) accept that as ok

Yes, I meant a non-story for the audience of Hacker News.

So deleting the files when not prompted to is correct behavior?

Files not recovered -> human in the wrong

Files recovered due to independent backup -> no problem

Is there any scenario in which you'd accept Claude's behavior as faulty here?

[delayed]

This is obviously bad model behavior (though I'd personally like to see the full transcript to see if there are any clues in there as to why it made the mistake.)

It's hardly new, though. We've been talking about models making stupid mistakes like this for years.

From <https://www.acm.org/code-of-ethics>:

  ACM Code of Ethics and Professional Conduct

  Computing professionals' actions change the world. To act responsibly, they
  should reflect upon the wider impacts of their work, consistently supporting
  the public good. 
  ...
  1.2 Avoid harm.
  In this document, "harm" means negative consequences, especially when those
  consequences are significant and unjust. Examples of harm include unjustified
  physical or mental injury, unjustified destruction or disclosure of
  information, and unjustified damage to property, reputation, and the
  environment. This list is not exhaustive.

"There were no fatalities and the victims are in stable condition."

Non-story!

I work at Anthropic. We were in touch with this professor the day he reported this issue. It’s very unfortunate and we’re glad everything’s been recovered. Claude clearly made a mistake here.

For anyone letting Claude work on their files, we recommend using auto mode. We made it the default in Claude Code, and it’s rolling out as the default in the Claude apps. I can’t say for sure if auto mode would’ve prevented this exact case, but this new default checks actions for risk before they run.

Auto mode as in auto destruct?

"Professor clicked delete on some files and later got them from the Trash Bin" is as interesting a story.

When my nephew was nine years old, he asked to use my little starter's pistol, a loud cap gun. His dad is a gun guy, so I asked my nephew if he had been taught how to properly handle a gun. Don't point it anyone, keep your finger off of the trigger, and so on. He solemnly said yes. I asked him to promise to treat it like a real gun, and he did. So I ceremoniously handed the pistol over. He immediately pointed it at my chest and pulled the trigger, then ran laughing out of the room, leaving me in shock from the very loud bang.

I treat AI like I now treat nine year olds with guns.

Why they never teach git in school?