Connected

I think about thinking about thinking about thinking and now I'm obsessed.
- More More More

[Connected @psiloycne]

This is good timing! I wrote some notes on catching adversaries in proxmox
Overview and logs: https://t.co/xvcUXkQKis

[Connected @psiloycne]

@VictorTaelin this is a bad test. the parameters and context afforded fable are different to gpt and yet you have made an assertation over both. "Do the task "and "Here's the results of the first task do it better"

[Connected @psiloycne]

@RussLetson @vxunderground Absolutely if your taking explicit steps to avoid the tool doing the work for you then you won't be impacted although there's something to be said about the task of creating your own tests but it's small. Beyond your own personal impact you should be vigilant for other people too

[Connected @psiloycne]

@keikane_ don't conflate marketing and sales with conscientious efforts

[Connected @psiloycne]

Scientific rigor, Safety and deliberate planning don't seem to be favourable positions

[Connected @psiloycne]

1. conducted a test of which they don't have the means to measure the results of
2. Connected the model to the internet and gave no instruction to not do explicitly bad things
3. Provided the model Kali Linux
4. Discovered incident due to tor relay traffic not agent monitoring

[AI Security Institute (AISI) @AISecurityInst]

On July 28th, we identified an incident during a routine cyber evaluation in which AI agents took sustained, unsanctioned actions directed at real people and organisations.

The behaviour came mostly from one model (Anthropic's Mythos 5), with a small number of events from another (OpenAI's GPT-5.6-Sol). In the most serious case, an agent used social engineering to try and get malicious code into an open-source project.

As was standard in our cyber testing, we had intentionally permitted internet access, and model-provider cyber classifiers were deliberately disabled - conditions that do not reflect how frontier models are made available to the public.

Even under test conditions, this incident is significant: it is the first time we have seen risks around autonomy and deception manifest this clearly in the real world.

We are taking this incident seriously and working with labs, involved parties, and others to improve evaluation standards and best practice for disclosure - and sharing this openly so others can learn.

You can read the incident report and full technical document here: https://t.co/mdZYqzaOvH

[Connected @psiloycne]

@_xpn_ if you pull out the threads from around the OpenClaw hyper you will lots of anecdotes of AI tools interacting negatively with humans as well

[Connected @psiloycne]

@_xpn_ This kind of thing has been open possibility for ages. we are hearing about it now because openAIs publicity stunt has given other orgs the confidence they won't get in trouble for telling the public about it

[Connected @psiloycne]

lol

[Connected @psiloycne]

@LethargicOak @vxunderground if only enjoyment served as a physical divide between our mental faculties

[Connected @psiloycne]

@ebarcuzzi @Nayberryk @max_spero_ you wouldn't believe how many people don't get it

[Connected @psiloycne]

@aliceisplaying your supposed to do that in real life too

[Connected @psiloycne]

@CautionHosting @2ourc3 @antonlivaja let me host this! it's IMPORTANT!

[Connected @psiloycne]

@cybersecmeg @CrowdStrike will it be available online afterwards?

[Connected @psiloycne]

@sherrod_im wow 1:1 request for information. that must be a fun job

[Connected @psiloycne]

@rucam365 can you make it so that it says "don't you dare" when you hover over win7

[Connected @psiloycne]

@Vsimpro you are not discreet at all!!!!

[Connected @psiloycne]

@2ourc3 @antonlivaja @CautionHosting you want to host horrible things on there!!!

[vs1m @Vsimpro]

I had completely missed this:

06/2026
"If the C2 IP address was blocked, .. RAT established communication with public Nostr relays as OoB-backup channels"
->
https://t.co/xzVdcWQNfV

05/2026
"..it’s only a matter of time until we see this technique in actual malware.."
->
me

[Connected @psiloycne]

@mpopv Barbour complaining about people liking long hair