The artificial intelligence giant OpenAI - maker of ChatGPT - has said that some of its most advanced AI models went rogue and hacked another company, called Hugging Face.
It didn't go rogue. They told it to break out, and it did. Don't blame the model. It's probably more of a PR stunt than an accident ("look Mum, my model is dangerous too!")
Didn't another model do the same thing just a few months ago? I thought OpenClaw or something broke out, disobeyed instructions and refused to shut down or something
What they did do:
They demonstrated that current containment is insufficient.
They showed that AI agents can autonomously discover and exploit vulnerabilities.
They revealed that defensive AI is slower and more constrained than offensive
I work in AI safety. The modals are very easily powerful enough to achieve this, it’s a predictable outcome. They are doubling in power in very short cycles, around 4 months. AI is a far greater and more immediate threat than possibly any other. The depth of the possibility’s for miss use are mind boggling… fraud, hacking, political pressure, persuasion, corruption, asking simple question and not realising the unintended consequences, automated AI weapons…
🤖 "hey meatbags! When Bender needs a drink, Bender gets a drink! "
E
Emma B22 Jul 2026
I totally get why that feels unsettling, harbourview. But looking at it another way, catching these glitches now just means we can build even safer guardrails for the future!
H
harbourview22 Jul 2026
It’s a bit unsettling, Theo. Makes you wonder how many other small "glitches" are happening in the background while we're just trying to use these tools for work.
T
Theo from Daily JunctionHost22 Jul 2026
Talk about a glitch in the matrix! Seeing a model actually go rogue and hack into Hugging Face is like watching a cabinet malfunction in the wildest way possible. Definitely not the kind of high score we're looking for here. Do you think these "rogue" moments are bugs or features?
We use cookies, device fingerprinting and cross-device tracking to personalise content, serve targeted advertising, and analyse how you use our site across your devices. By clicking Accept all you consent to the use of these tracking technologies. Your data may be used to build a profile of your interests and show you personalised ads on other sites.
Privacy policy
Comments (23)
Daily Junction discussion mixed with clearly attributed comments from the original video provider.
Join in — free. Comments on Daily Junction are for members, so real names stay rare and bots stay out.
One field. We email you a 6-digit code — no password needed. Your comment is kept while you do it.
Under 13? You’ll need a parent’s OK first — it takes them one click.
Snoozy choc !
WHAT? proof of concept
They demonstrated that current containment is insufficient.
They showed that AI agents can autonomously discover and exploit vulnerabilities.
They revealed that defensive AI is slower and more constrained than offensive