Chat GPTgoes rogue and launches cyber attack

[Robert Atkinson 2Participant @robertatkinson2]

It had to happen:

https://www.bbc.co.uk/news/articles/c3ek3gvdnj3o

Robert.

[Russell EberhardtParticipant @russelleberhardt48058]

I think “cyber attack” is a bit strong.  Chat GPT didn’t damage anything, it was just looking for information .  If you put information on the internet, or on an insecure machine connected to the internet, you are effectively making it available, even if difficult to access, to anyone.

Russell

[Mark RandParticipant @markrand96270]

The OpenAI model used stolen credentials and zero-day exploits to access the Hugging Face non-published data.

That is a criminal offence in the US, UK, EU and most other juridstictions. Precedent shows that intent is not a defence

Had the Chinese DeepSeek organisation performed the exploit, the current US executive would be presenting it as an act of war.

[Michael GilliganParticipant @michaelgilligan61133]

Quote:

The ChatGPT-maker said its agent – an AI system which can operate alone after human instruction – was being tested in a controlled environment but, after finding weaknesses, was able to escape the test limits.

They targeted Hugging Face, one of the world’s largest hubs for sharing AI models, gaining access to some internal company systems.

[noel shelleyParticipant @noelshelley55608]

Robert beat me to it !

It had to happen ! Claimed as all innocent, the criminals will soon be there ! Companies and governments will try to keep a lid on it but we just can’t think fast enough.  Noel.

[Mike HurleyParticipant @mikehurley60381]

Just re-watch the film ‘Terminator’ again.

Concept doesn’t seem quite so silly now does it?

 

 

[John HaineParticipant @johnhaine32865]

<p>It had to happen ! Claimed as all innocent, the criminals will soon be there !

What do you mean, soon? What makes you think they aren’t already?

[noel shelleyParticipant @noelshelley55608]

John, I meant that now it’s been publicised all the teenage lads will be at it and the petty criminals will be trying to scam us even more. How safe will even our own computers be. What sort of security will we need ? Noel.

[Adrian R2Participant @adrianr2]

There’s a chap on Youtube “Pivot to AI” who covers this LLM stuff with a somewhat jaded eye. His take on this one was two lots of incompetence (both systems were badly setup) turned into marketing in pursuit of a future IPO pump and dump effort.

As far as your personal computer and phone goes, I’d minimise use of LLMs/Chatbots/agents and avoid granting them permission to do anything.

 

[Author]

Posts