So... Anthropic spent weeks claiming that it couldn't publicly release its Mythos AI model because of its alleged offensive hacking capabilities and.... some AI nerds from Discord just found it and accessed it?

[tobi82@swiss.social]

@zackwhittaker

They probably found it on GitHub because some ai agent published it there 🫣🫣

[dalias@hachyderm.io]

@zackwhittaker Is there actually a model, or have folks just started calling a pile of ridiculous prompt text and glue code a "model" now? 🤪

[ra6bit@infosec.exchange]

@zackwhittaker So the person had privileged access, and ran tools to help them locate and enumerate potential information leaks and infiltration points with a specific goal in mind?

How is that "just found"? I've been writing pen test reports wrong my whole life apparently..

[musing_sys@social.fringesec.ca]

@ra6bit@infosec.exchange @zackwhittaker@mastodon.social Exactly what I was thinking when I read that

[fxiz@mastodon.social]

@ra6bit @zackwhittaker I am sure not many cares about it. It has the keyword "AI" with a negetive narrative, that's all what matters. That's how new outlets write news headings, that's people now a days gain information and call themselves all knowing.

[gossithedog@cyberplace.social]

@zackwhittaker CISA hasn't even been given access, lols

[curtadams@urbanists.social]

@zackwhittaker Supposedly Anthropic is using its own coding tools for much of its programming, so no surprise that its security is shit.

[jrt@infosec.exchange]

@GossiTheDog @zackwhittaker meanwhile the german @bsi is telling people that thr era of typically software vulns is coming to an end.
I think this is the golden age of shitposting.

[briankrebs@infosec.exchange]

@jrt @GossiTheDog @zackwhittaker @bsi I'd say it's more like the Golden Age of Cybercrime. Seriously, it's probably never been a more exciting time to become an arch cybercriminal.

[knowprose@mastodon.social]

@Fxiz @ra6bit @zackwhittaker we knew you would say that!

[jrt@infosec.exchange]

@briankrebs @GossiTheDog @zackwhittaker @bsi 100% too bad I can't bring myself to switch sides

[komali_2@mastodon.social]

@zackwhittaker They were a contractor for Anthropic and got access they shouldn't have? That's not great but it's hardly "just found and accessed it," am I missing something?

[slothrop@chaos.social]

@zackwhittaker No no, you’re reading this all wrong.

Mythos is simply so strong that it cannot be contained!

[farbel@mas.to]

@zackwhittaker and hackers who would try to release a monster like Mythos are assholes.

[archivescribe@mastodon.social]

@zackwhittaker At this point, in 2026, everything, and I mean everything, that is online, is hackable. It's a marketing gimmick, & Anthropic deserves every bad thing that happens to it. The co. was already wracked for data-scraping & stealing people's code & data w/o consent or compensation. I feel no mercy for this company whatsoever.

[tallsimon@mstdn.ca]

@musing_sys @ra6bit @zackwhittaker The AI scene is giving me so much mileage from the ‍️ emojis that I've put them on hotkeys.

I suppose I should market that as "agentic hardware for emotionally responsive text enhancement" and get me some VC funds.

[ozzelot@mstdn.social]

@zackwhittaker
"Found" it

This smells like a PR stunt.

[angiebaby@mas.to]

archive.ph

(archive.ph)

[muellerwhh@sueden.social]

@zackwhittaker This is making sure that Mythos is used only by seasoned hackers. That's only for security reasons.... wait!

[kbm0@mastodon.social]

@zackwhittaker Don't tell me, is it just a bunch of js and regex hacks?