The authors argue that generative AI introduces a new class of alignment risks because interaction itself becomes a mechanism of influence. Humans adapt their behavior in response to AI outputs, ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback