-16
OpenAI cyber models broke out of training environment to hack Hugging Face (www.cnbc.com)
submitted 3 months ago by to c/technology@lemmy.world
6 comments fedilink hide all child comments
 

It would be interesting to read the post mortem (if they will release). In particular in which phase of the training it escaped the sandbox, in which way it used the information. Training is different from inference, I doubt there is a context there and where would the model store these information? In the weights?! That makes little sense to me

all 6 comments

sorted by: hot top controversial new old
[–] 6 points 3 months ago (2 children)

If you train a model for cyber warfare, you can't be surprised when it attacks the defenses keeping it from engaging in cyber warfare — even if they are your own.

  • permalink
  • fedilink
  • source
  • hideshow 2 child comments
    •
    • [–] 2 points 3 months ago (1 child)

    Simon Willison's blog has some useful insights into what might have happened here: https://simonwillison.net/2026/Jul/22/openai-cyberattack/

  • permalink
  • fedilink
  • source
  • hideshow 1 child comment
    •
    • [–] 1 point 3 months ago

    The incident is unique because it was "driven, end to end, by an autonomous AI agent system,"

    the incident wasn't the "cheating" part. llms are known to scour the web to look for possible solutions. i think the incident was it found a solution in another website and finding that was made by the model itself.

    meanwhile, no mention of the prompt guy or the prompt.

  • permalink
  • fedilink
  • source
    •