| name | 2028-ai-leadership-scenarios |
| category | ai_collection |
| description | Methodology from Anthropic policy analysis examining possible trajectories for US-China AI competition by 2028 — focusing on compute advantage, export controls, distillation attacks, and two scenarios for democratic vs. authoritarian AI leadership. |
| tags | ["anthropic","policy","ai-leadership","scenarios","us-china"] |
2028: Two Scenarios for Global AI Leadership
Methodology from Anthropic policy analysis (May 14, 2026) examining possible trajectories for US-China AI competition by 2028.
Core Thesis
It's essential that the US and its allies stay ahead of authoritarian governments in AI development. AI will soon become powerful enough to be used for repression at unprecedented scale and alter the balance of power among nations. The window to set competitive conditions is limited.
Key Factors
Compute Advantage
- Most critical ingredient for frontier AI development
- Currently held by US through American chip innovation
- Tightened via bipartisan US export controls on advanced compute
How China Has Stayed Close
- World-class AI talent
- Loopholes in US export controls
- Large-scale distillation attacks: Illicitly extracting US model innovations by harvesting output patterns
Two Scenarios for 2028
Scenario 1: Democratic Leadership (Preferable)
- US successfully defends compute advantage
- Policymakers tighten export controls and close loopholes
- China's distillation attacks disrupted
- Democracies accelerate AI adoption
- Democracies set rules and norms around AI
- Likely 12-24 month lead in frontier capabilities locked in
- Safety engagement with China more feasible
Scenario 2: Authoritarian Convergence
- US chooses not to act or fails to act
- Loopholes in CCP's access to compute remain open
- Chinese AI firms catch up and potentially overtake
- AI norms and rules shaped by authoritarian regimes
- Best models enable automated repression at scale
- "No solace that authoritarian triumph happened on American compute"
Four Fronts of Competition
- Compute access and chip supply chains
- Model intelligence and distillation defense
- AI adoption rate across democratic economies
- Global distribution and norm-setting
Policy Recommendations
- Tighten export controls on advanced compute to PRC labs
- Disrupt distillation attacks targeting US models
- Accelerate democracies' AI adoption
- Prepare for transformative AI systems arriving by 2028
Activation
AI leadership, US-China, AI policy, scenarios, global AI competition, 2028, export controls, compute advantage, distillation attacks
