As AI agents grow more autonomous - booking travel, executing transactions, acting on our behalf across digital systems - a fundamental question emerges: do we treat them as quasi-human entities deserving of trust, identity, and even rights, or as sophisticated programs to be controlled, audited, and constrained? The answer is far from academic. How we frame AI agents shapes the security models we build, the identity architectures we deploy, the liability frameworks we establish, and the attack surfaces we expose. This panel explores the tension between these two worldviews, interrogating the assumptions behind our current approaches and asking what is truly at stake as agents become more capable, more embedded, and harder to distinguish from the humans they represent.
