Current law struggles with non-deterministic errors because traditional negligence requires proving a human failed to meet a standard of care. When an AI makes an unpredictable decision, finding a specific point of human failure becomes difficult. We likely need to move away from fault-based liability toward a strict liability model for high-risk agents.
Under strict liability, the entity that profits from the AI's operation bears the financial burden of its errors, regardless of whether they intended the mistake. This creates a natural incentive for companies to implement rigorous guardrails and kill-switches. Think of it like owning a wild animal; if it escapes and causes damage, the owner pays.
Another option involves mandatory insurance pools. Developers could contribute to a collective fund designed to compensate victims of
