[Waterluvian]

Just like using an AI model, you can’t actually know for sure that it won’t do anything malicious with what interfaces you give it access to. You just have to trust it.

[sahilagarwal]

Isn't that true for every piece of software you use? Reflections on trusting trust (https://www.cs.cmu.edu/~rdriley/487/papers/Thompson_1984_Ref...) was one of the formative papers of my software engineering career.

[dkga]

Well, you can at least check if there is network traffic to AWS or something similar.

[yurishimo]

But wouldn't that look the same as actually querying the model? Or am I missing the joke?

[Waterluvian]

There’s always ways to mitigate malicious behaviour once it’s already happening.