Who is liable for AI agents? Paper proposes 'A-corp' legal framework
scychan_brains · x · 2026-08-12
As AI agents proliferate and autonomously take billions of actions, harm is inevitable. Because AIs lack physical bodies and can copy, merge, or vanish at will, the primary legal challenge when things go wrong is identifying 'Which AI did it?'
To solve this accountability dilemma, the paper proposes two types of identity: 'thin' identification, which ties AI actions to a human principal, and 'thick' identification, which distinguishes between millions of discrete AI entities. The authors introduce the 'Algorithmic Corporation' (A-corp)—a legal-fictional entity owned by humans but run by AIs. This framework ties AI actions to human owners for accountability and allows the legal system to restrict financial or compute resources to curb unlawful behavior.
More from AGI Musings
- User Wonders Why Models Lack Deep Understanding of Their Own Trajectory — willdepue · 2026-08-12
- Mathematician Admits Math Community Concerns Will Be Quaint Amid Massive AI Upheaval — littmath · 2026-08-12
- Mathematician Predicts AI Autoformalization Will Repair Literature but Disconnect Ideas — littmath · 2026-08-12
- Mathematician Warns Current Institutions Will Fail to Train Talent as AI Disrupts Incentives — littmath · 2026-08-12
- AI Creates Prisoner's Dilemma in Math: Scholars Refuse to Discuss Unpublished Work — littmath · 2026-08-12
- Data Reveals AI Impact: MathOverflow Q&A Plummets Since Early 2025 — littmath · 2026-08-12