Ohhh whoa....
https://www.anthropic.com/research/discovering-cryptographic-weaknesses
#cryptography #ClaudeMythos
Remote
joy larkin 🌺✨
@joy@mastodon.social
mastodon 4.8.0-nightly.2026-10-06👩🏻💻Technologist. Leader, builder, and advisor in Silicon Valley. #AI, #Infra, #OSS enthusiast. Single mom with an amazing daughter. A part of my heart is in 🇪🇺. #IndieMusic fan. NJ/Philly✈️SF. ❤️✨
#Tech #OpenSource #SF #SFBA #SFBay #SanFrancisco #fedi22
#RSS
@joy.rss@mastodon.social
https://cleverhack.com/feed
*Side Projects
https://cleverhack.com/frontier-ai-lab-tracker
https://cleverhack.com/ai-coding-landscape
https://github.com/joylarkin/Awesome-AI-Market-Maps
0 Followers
0 Following
24 Posts
Joined October 06, 2016
cleverhack dot com:
GitHub:
Twitter:
Pixey:
Replying to
@andrewnez@mastodon.social
This repo I'm maintaining about OpenClaw security headlines might help with your research
https://github.com/joylarkin/openclaw-security-news
Replying to
just make sure the devs are comfy
My Joker origin story in a past life or two...
Anthropic's security bug bounty program is now available on Hacker One.
While hearing the news of the reintroduction of gray wolves in Sequoia National Park, I was reminded of my favorite piece of outdoor writing.
https://www.outsideonline.com/outdoor-adventure/environment/adventure-very-old-man-wolf/
Oh whoa, someone wrote up a whole developer guide to my AI Market Maps repo.
Oh whoa, Massie is projected to lose tonight...
Replying to
@bplein@bvp.me
If it wasn't coming from Anthropic (or DeepMind for that matter), I'd agree with you). All I can offer is my perspective and what I'm hearing from being up closer to this all than most.
Replying to
@Viss@mastodon.social
Yeah, the one thing that is new is the number of machines (4000) that were compromised.
Replying to
@briankrebs@infosec.exchange One thing about this ever persistent social graph building... I'm just more (or less) shocked by the number of long-ago exes who apparently still have my number in their contact lists.
Replying to
@metabolist@mastodon.social Thanks for your hard work, and please take care of yourself. ❤️
Hey @Viss@mastodon.social did you see that someone prompt injected NemoClaw (the NVIDIA stack for OpenClaw)?
https://bdtechtalks.substack.com/p/how-prompt-injection-broke-nvidias
June Gloom (sorry, May) so dark in the South Bay that the outdoor lights are still on at 7am this morning. #sfba
Replying to
@bplein@bvp.me
The Red Team report is here --> https://red.anthropic.com/2026/mythos-preview/
And here is the Model System Card --> https://www-cdn.anthropic.com/8b8380204f74670be75e81c820ca8dda846ab289.pdf
Replying to
@bplein@bvp.me
They are from other teams at Anthropic - their security and their alignment teams.
The findings haven't been disputed, most of the critiques I've seen from others in the industry is the "how they got there."
Replying to
@Sikorsky78@infosec.exchange
It is more that we've now gotten to the point where there are systems able to do the math at a faster rate than before to break cryptography.
Replying to
@Viss@mastodon.social
This is old news. :) What I want to know is what other threat actors are using OpenClaw for c2.