Agent permissions resemble human permissions: think delegation, not processes
charles_irl · x · 2026-09-10
- Researcher charlesirl argues that the permissions we want to give agents substantively resemble human permissions, because agents are trained to mimic humans and pursue human goals.
- The right frame is therefore delegation of the agent creator's authority, access, and identity — not merely issuing new permissions.
- Responding to the objection that processes/threads and containers already handle permissions, he notes the hard part is defining the nature of the delegation: what exactly is delegated, and whether and by whom it is alienable.
Related event: Permission Delegation Between Agents Remains an Overlooked Security Gap(2 posts)→
More from AGI Musings
- Richard Socher: goal setting is the key to controlling superintelligent AI — RichardSocher · 2026-09-11
- Defending AI risk concerns: the marketing-conspiracy theories don't hold up — AndyMasley · 2026-09-11
- AI to cure all diseases in 5 years then end humanity, quip sparks mockery — MaxUnfried · 2026-09-11
- Is quantifying uncertainty arbitrary? A Bayesian ideology debate — NathanpmYoung · 2026-09-11
- Zvi disputes claim that AI lost a live debate against conservatives in Dallas — TheZvi · 2026-09-11
- Matt Turck talks with Richard Socher on recursive self-improvement and AI-driven science — mattturck · 2026-09-11