Policy proposal expressed as a scenario
Vardark interpretation of AI 2040: Plan A.
07 / GOVERNANCE-LED TRANSITION
AI 2040: Plan A
An international agreement changes the incentives for developing powerful AI, combining research transparency, verification and controlled scaling. This is a proposed route to a flourishing future, not a prediction or Vardark policy endorsement.
- VARDARK STATUS
- Not assessed
- CONTENT REVIEWED
More context
The authors propose an international agreement that replaces a secretive race with transparent research and coordinated development. Multiple countries and companies catch up while scaling is deliberately constrained.
How the pathway unfolds
Agreement and transparency
The scenario places a US–China agreement in 2029 and automation of AI research in 2030. Shared research and verification support coordinated limits.
Controlled scaling
Systems reach top-human-expert capability, pause in 2035, then scale to superintelligence in 2040. The proposal includes mutually assured compute destruction as an enforcement mechanism.
What it assumes
- Major powers reach and sustain an enforceable agreement despite incentives to defect.
- Verification and security can constrain covert projects, while research makes later scaling controllable.
Where intervention could matter
Conditional intervention points in this account, not proven safeguards.
- Negotiate coordinated limits and research transparency before a destabilizing race.
- Make verification, security and alignment progress conditions for further scaling.
Questions that could weaken this account
Vardark questions for future assessment. These do not establish that counterevidence has been observed.
- Could covert development or ambiguous capability thresholds make compliance impossible to verify?
- Would enforcement create unacceptable conflict risks or concentrate power despite the proposal’s aims?
What we can assess today
The Observatory does not assess treaty feasibility, verification effectiveness or retained human control. Existing compute, energy and research measurements provide context, not proof that this governance plan works. These are stress-test questions, not findings against the proposal.
Dimensions and current coverage
The states are ordinal. They are not percentages and are not added together.
International coordination
No direct evidence link
Verification
No direct evidence link
Retained human control
No direct evidence link