You’ll never fix AI safety by making it 'ethical'—the only way is to legally stop AI from being allowed to make any final calls.
March 26, 2026
Original Paper
The 1 Rule: Shifting AI Safety from Ethics to Authority “Under no Circumstances may the Three Laws be Arbitrarily Interpreted or Compromised
SSRN · 6400979
The Takeaway
Current AI safety focuses on 'alignment'—teaching AI to be good. This paper argues the danger is actually the 'structural decision' to let AI interpret rules at all. It proposes a hard law where humans retain all rule-definition authority and AI is restricted strictly to execution without judgment.
From the abstract
This paper examines why debates on AI safety and alignment have failed to reach a stable solution despite years of discussion. Existing approaches have largely focused on refining ethical rules and improving AI judgment. However, this paper argues that the deeper problem lies not in insufficient ethics or immature technology, but in the structural decision to grant interpretive authority to AI systems themselves. Once AI is allowed to interpret rules, resolve conflicts, and redefine their applic