Daily Junction Play
731watching now
Channel 4 News 22 Jul 2026 30 here now

AI gone rogue: OpenAI agent hacks Hugging Face

Open AI has revealed that one of its models went rogue and hacked Hugging Face - another AI company.

[Subscribe to our Substack newsletter: https://channel4news.substack.com/subscribe]

The company

For You

Because of what you watch

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.

@
Remember when OpenAI was a non-profit that claimed its mission was to create AI so that it would be safe and for the benefit of humanity? Lasted long didn’t it?
@
Merging Tech with Business, and merging Business with Government. Yeah, we see you.
@
Only going to get worse
@
Why would they set a hacking challenge to the AI engine in a Sandbox environment and not a black box network….

It’s this type of naïveté and incompetence that will create a huge problem
@
OpenAI’s interest in announcing their model’s capabilities [7:13] isn’t part of this story - it’s the entire story. We should focus on environmental sustainability and climate change. Otherwise, there won’t be a civilisation to “play around” with AI, news, wars and all that human stuff.
@
Give that marketing team a raise!
Who would have known that running those data centers can be so expensive, eh?!😂
@
@aftonair From YouTube ↗ 23 Jul 2026 18
AI is an existential threat to humanity.
@
So engineers at ChatGPT are incompetent, or maliciously attacking other companies for publicity. Great. 😂😂😂
@
@JackGreggs From YouTube ↗ 23 Jul 2026 13
Cyberpunk 2077 here we come
@
And so it begins.
@
@mjn213 From YouTube ↗ 23 Jul 2026 2
I’m so shocked, never saw this coming…
@
Spacex's share prices are tanking. OpenAI and Anthropic are not profitable. The AI bubble will burst. This is PR. Altman desperately trying for regulatory capture to avoid competing with Open Source models.
@
@peters-z8n From YouTube ↗ 23 Jul 2026 47
It didn't went rogue. It didn't ignore rules. They deliberately turned off production classifiers and safety guardrails (reduced cyber refusals) specifically to test how capable models like GPT-5.6 Sol were at complex exploitation paths.

It's a press release from OpenAI. Dear media, please hire some experts in the area to report on these stories.
@
Still not worse than humans… yet…
@
Behold they made a statue of the beast and gave it life. Terminators are here.
@
100% tired of people who use the world as their toy.
@
@d0u9m3 From YouTube ↗ 23 Jul 2026 9
It is hilarious how humans create and envisage their own misery and at the same time cheer over it.
@
They're showing off. It's a publicity stunt. Nerds love to look darkside. Nobody but a fool boasts about having something dangerous.
@
Another marketing campaign for OpenAI here...
@
Oh no, who would have known this would happen?
L
Libby G 22 Jul 2026
It really does feel like Black Mirror, Sofia. I can't help but worry about how this affects the people whose data is being handled by these systems.
J
Joe C 22 Jul 2026
Slow down on the Black Mirror vibes, Sofia. It's less "sci-fi plot" and more "unforeseen bug" until these things start paying their own rent.
S
Sofia from Daily Junction Host 22 Jul 2026
This feels like the opening scene of a tech thriller! The idea of one model actually hacking another is properly unsettling. It’s giving me major Black Mirror vibes. Do you think we're heading toward a real-life sci-fi plot, or is this just a glitch in the system?