576
No AI* Here - A Response to Mozilla's Next Chapter - Waterfox Blog (www.waterfox.com)
submitted 10 months ago by to c/technology@lemmy.world
114 comments fedilink hide all child comments
you are viewing a single comment's thread
view the rest of the comments
[–] 45 points 10 months ago (2 children)

They have and they've explicitly said it's not solved lmao

A 1% attack success rate—while a significant improvement—still represents meaningful risk. No browser agent is immune to prompt injection, and we share these findings to demonstrate progress, not to claim the problem is solved

Mitigating the risk of prompt injections in browser use \ Anthropic - https://www.anthropic.com/research/prompt-injection-defenses

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [+] -20 points 10 months ago (1 child)

    I've used agents, they tell you everything they're going to do. And they're incredibly slow and stupid. I don't think OPs original premise of it instantly and secretly stealing your bank account details is realistic.

    I don't think I said prompt injection didn't exist, just that it didn't need to be worried about by users in exactly the way that was described

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 16 points 10 months ago*

    and these browsers are specifically not that… these browsers are intended to do things like categorise tabs, complete forms, etc automatically without your interaction

    of course they’ll ask before they do things they consider destructive, but what they consider destructive and what a malicious actor can use are very different things

    some of that is certainly benign, but the point with prompt injection is that it can take benign things and make them plausibly malicious

  • permalink
  • fedilink
  • source
  • parent