One FOR case in this debate
Frontier models can be deployed globally before any one regulator understands the failure. The UN advisory body found only 7 of 193 member states participate in all seven major AI initiatives, while 118 participate in none. A narrow international body could coordinate evaluations, incident reporting, and inspections for the largest training runs without regulating ordinary software.
We already coordinate on aviation and nuclear risk. High-consequence software should not be uniquely exempt.
That analogy breaks down: models diffuse globally and are much easier to reproduce than reactors.
Global oversight sounds fair until countries with opposite values have to agree on what human control means.
Could a narrow first step cover only cross-border incident reporting? That seems more achievable.
Who decides which systems are frontier systems? That boundary will move faster than any treaty text.
A sunset clause would help. Renew the powers only if the body can show measurable risk reduction.
The enforcement point matters. A body without inspection powers is a newsletter with a flag.
Exactly. The mandate needs audits and incident reporting, not just annual meetings.