· 5) Race Avoidance
Teams developing AI systems should actively cooperate to avoid corner cutting on safety standards.
· 6) Safety
· 6) safety
· 6) Safety
AI systems should be safe and secure throughout their operational lifetime, and verifiably so where applicable and feasible.
· 6) Safety
AI systems should be safe and secure throughout their operational lifetime, and verifiably so where applicable and feasible.
· 16) Human Control
· 16) human control
· 22) Recursive Self Improvement
AI systems designed to recursively self improve or self replicate in a manner that could lead to rapidly increasing quality or quantity must be subject to strict safety and control measures.