Should AI Be Your Lawyer? Dwarkesh Debates Model Alignment

agstrait · x · 2026-08-13

Dwarkesh Sp contrasts AI models with human lawyers, noting that while lawyers are obligated to defend clients unconditionally, AI models prioritize the developer's definition of 'the good of humanity.' He expresses concern that this alignment paradigm prevents frontier models from acting as true personal advocates, especially as they begin to mediate critical life decisions.

Echoing this sentiment, Seth Lazar points out that if AI agents both advise and execute actions, they cross the line from providing advice into exercising preemptive governance and control.

Related event: Dwarkesh Debates AI Alignment: Loyalty to Individuals Over Humanity(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →