Using AI as political force, disinformation tool, and in general for speeding up science researches is not new, and personally see that quite impressive for last decade. But enterprise grade models can’t built in garage, it requires many experienced specialists in domain, lot of specialised devices and enegry. The question is - how we can restrict it from using by anyone?

From one side I’d like to have access to hight quality models for autio/face/text and code generation that can be run without expensive subscriptions by tokens amount. From other side I’d like to avoid that any random person would use it maliciously. So, what is the answer?

I think there two options: make it only available by subscription (which doesn’t protect from unlicensed services) or make it fully traceable by generating hidden footprints. Which quite technically challengeable, but no impossible. And like with guns, I guess we can go both ways.