AI slop classic, somebody files a bug report for a vuln, doesn’t disclose they’ve used AI when asked, argues in the comments - and includes his AI prompt in the paste.

https://hackerone.com/reports/3230082

@GossiTheDog@cyberplace.social Did it not notice the obvious error in the cast which should have been (char *) not (char)? Any compiler would have caught that.
@nowster @GossiTheDog It's not Watcom, it's Eliza.
×

AI slop classic, somebody files a bug report for a vuln, doesn’t disclose they’ve used AI when asked, argues in the comments - and includes his AI prompt in the paste.

https://hackerone.com/reports/3230082

I think the open source movement is in serious trouble due to generative AI, as maintainers are going to get flooded with torrents of bad code and analysis the longer the AI hype train continues.

It’s exhausting and unsustainable if an entire generation are introduced to not knowing or caring about what they are doing.

@GossiTheDog that's what prompted me to write a Human Content Policy.

https://github.com/gsuberland/altium_js/blob/main/HUMAN.md

altium_js/HUMAN.md at main · gsuberland/altium_js

altium.js - Altium SchDoc parser and renderer, in the browser. - gsuberland/altium_js

GitHub
@gsuberland @GossiTheDog I know this isn’t the point but this is the first time I’ve seen your project and it’s something that could be quite useful
@GossiTheDog We can entrench to some degree, but it will make development more brittle, and it runs against cultural pressures towards figuring out more open social contracts. Pretty tricky, indeed.
@GossiTheDog I foresee projects will end up just banning vuln reports from accounts who have not been vetted.
@GossiTheDog Makes me wonder what motivates them. The unlikely prospect of five minutes of internet fame?
@GossiTheDog time to automatically reject every report that doesn't include at least two typos 😌
@GossiTheDog Open source is dead but its because of AWS, GCE, and Azure as well as private equity.

@GossiTheDog I’ve seen well meaning contributors *champion* AI for the FOSS projects they work with… and for what? Most FOSS doesn’t *need* productivity. And it certainly doesn’t need huge black spots of code with no real owners because one person shovelled in more features than they could ever reasonably understand or support. 🫠

(I have written spades on this but I just don’t have the spoons for this battle)

@GossiTheDog If you put a few "woke" words in your project description or files then most of the AI crap runs away.

@etchedpixels

Add a commented-out "fuck" on every other line

@GossiTheDog

@etchedpixels @GossiTheDog please elaborate :) I am most interested :)
@GossiTheDog we're about to understand how the dark ages happened
@GossiTheDog sure but it's not limited to foss
@Laukidh the one thing the article doesn’t state is the success rate of the AI at the point of human validation. If you have a team of skilled analysts dropping 90% of the AI’s output as junk then how successful is it really?
@GossiTheDog @athos We are going to get unintentional Skynet.
That slop squared, no?
@GossiTheDog
I'd call that slop squared...
@GossiTheDog@cyberplace.social Did it not notice the obvious error in the cast which should have been (char *) not (char)? Any compiler would have caught that.
@GossiTheDog don't argue with them, just demand a PoC. If they can't provide a PoC they're full of shit
I think a video PoC could be a requirement at this point. Certain vulns could have specific minimum reporting standards to be included in the videos. Bonus points for using debuggers and showing repo hash verification for projects that have source code available.
@GossiTheDog This is hilariously not funny🤦🏽‍♀️🤦🏽‍♀️. If that makes sense.