alt.hn

4/22/2026 at 11:48:58 AM

Mythos Falls into the Wrong Hands

 https://www.theverge.com/ai-artificial-intelligence/916501/anthropic-mythos-unauthorized-users-access-security

by etothet

4/22/2026 at 12:29:27 PM

I don't feel like there's nearly enough evidence here to know whether the claims are accurate. The group claimed they accessed it and used screenshots as evidence, whereas Anthropic says that they are looking into it but haven't found any evidence of it so far. I don't feel like screenshots are particularly strong evidence, but I also wouldn't expect Anthropic to admit to it if they did realize it happened either.

by saghm

4/22/2026 at 7:48:49 PM

Nor has any interesting information been provided by third parties in reviewing the claims made by anthropic about mythos. Almost everything I've seen is some version of "X has (inquired|reviewed|previewed) Mythos" which is indistinguishable from the new norm of 99% clout-seeking as advertisement in journalism.

by cyanydeez

4/22/2026 at 12:47:05 PM

As other mentioned, there is zero evidence of that, but of the company who claims did something so smart cannot use it to protect itself, it seems it is worthless.

Unless this is all lies to make people think Skynet is on the loose. And by now, anthropic and other aislopware are full of lies

by motbus3

4/22/2026 at 1:05:31 PM

At this point, this all feels like marketing, people getting paid left and right to make all kinds of claims about this model as long as it stays in the spotlight.

by overflowy

4/22/2026 at 1:50:25 PM

Yup seems like it’s working. I have no idea what “next” model is for any other company.

by derwiki

4/22/2026 at 2:27:12 PM

This feels like a PR stunt by Anthropic. We understand, Mythos is a great model. But please stop unneeded publicity for the sake of IPO.

by drewfax

4/22/2026 at 12:07:10 PM

"This is the one thing we didn't want to happen"

by oh_fiddlesticks