AI Capability Stratification
A Framework for the Future Distribution of AI Power
- Wu, Shaoyuan
Global AI Governance and Policy Research Center, EPINOVA LLC
https://orcid.org/0009-0008-0660-8232
Description
This policy brief proposes AI Capability Stratification as a structural framework for understanding the future distribution of AI power. It argues that public-facing AI represents only the visible surface of a broader capability structure, while deeper forms of AI power emerge through institutional embedding, proprietary knowledge, operational control, resource authority, real-world interfaces, and differentiated rule conditions.
Abstract
The future development of artificial intelligence should not be understood as a linear race among public-facing large models. As AI systems become embedded in institutions, workflows, operational systems, proprietary knowledge environments, resource-allocation mechanisms, and differentiated rule conditions, AI capability is likely to form a stratified structure. This policy brief proposes AI Capability Stratification as a structural framework for understanding the future distribution of artificial intelligence power. The framework shifts attention from model performance alone to the broader conditions under which AI capability is accessed, embedded, authorized, optimized, and connected to real-world systems. The model divides AI capability into three broad domains: Surface Intelligence Domain, Deep Intelligence Domain, and Dark-Domain Intelligence. These domains can be further divided into seven analytical layers: Public Interface Intelligence, Institutional Embedded Intelligence, Operational Control Intelligence, Proprietary Epistemic Intelligence, Objective-Specific Optimized Intelligence, High-Authority Resource Intelligence, and Boundary-Condition Intelligence. The central thesis is straightforward: Public AI determines visible access; Deep AI determines professional productivity; and Dark-domain AI determines how AI capability is linked to concentrated authority, resource command, real-world consequence, and differentiated rule conditions. This framework does not claim that deeper AI layers are always more intelligent, more advanced, or more dangerous. Rather, it argues that AI power depends on the interaction between model capability, data quality, access conditions, institutional embedding, resource authority, real-world interfaces, and rule space. The future distribution of AI power will therefore not be determined only by which model is the strongest. It will also be determined by who can use AI, what AI can access, what resources AI can influence, what systems AI can affect, and under what rules AI is allowed to operate.
Files
Mobile browsers may not display embedded PDF previews reliably. Open the PDF directly for the best reading experience.
Open PDF Download PDF| Name | Type | |
|---|---|---|
| AI Capability Stratification A Framework for the Future Distribution of AI Power.pdf Full-text PDF of the policy brief | application/pdf | Download |
Keywords
- AI capability stratification
- AI power
- Surface Intelligence
- Deep Intelligence
- Dark-Domain Intelligence
- Public Interface Intelligence
- Institutional Embedded Intelligence
- Operational Control Intelligence
- Proprietary Epistemic Intelligence
- Objective-Specific Optimized Intelligence
- High-Authority Resource Intelligence
- Boundary-Condition Intelligence
- AI governance
- AI systems
- model capability
- data quality
- resource authority
- real-world interfaces
- rule space
- institutional embedding
- AGI
- foundation models
- frontier AI
- AI regulation
- AI risk management
- EPINOVA
Subjects
- Artificial intelligence
- AI governance
- AI policy
- Technology policy
- Strategic studies
- Public policy
- Institutional governance
- AI safety
- AI regulation
- Foundation models
- Frontier AI
- Risk management
- Security studies
- Information systems
- Digital governance
Recommended citation
Wu, Shaoyuan (2026), AI Capability Stratification: A Framework for the Future Distribution of AI Power, Policy Brief No. EPINOVA–2026–PB–59, Global AI Governance and Policy Research Center, EPINOVA LLC. https://doi.org/10.67037/epinova.pb.2026.059.
APA citation
Wu, S. (2026). AI capability stratification: A framework for the future distribution of AI power. EPINOVA Policy Brief Series, EPINOVA-PB-2026-059. Global AI Governance and Policy Research Center, EPINOVA LLC. https://doi.org/10.67037/epinova.pb.2026.059.
Alternate identifiers
| Scheme | Identifier | Description |
|---|---|---|
| URL | https://epinova.org/policy-brief-1 | Official EPINOVA publication page |
| EPINOVA policy brief number | EPINOVA–2026–PB–59 | Policy brief number printed in the PDF |
| File name | AI Capability Stratification A Framework for the Future Distribution of AI Power.pdf | Source PDF file name |
| Short title | AI Capability Stratification | Short form of the policy brief title |
| Analytical concept | Surface Intelligence Domain | Visible public-facing layer of AI capability in the framework |
| Analytical concept | Deep Intelligence Domain | Permissioned, professional, organizational, and embedded domain of AI capability |
| Analytical concept | Dark-Domain Intelligence | Restricted-visibility, high-threshold, high-resource, and high-consequence domain of AI capability |
| Analytical concept | Seven Layers of AI Capability | Seven-layer stratification model developed in the policy brief |
Related works
| Relation | Identifier | Type | Description |
|---|---|---|---|
| IsPartOf | https://epinova.org/policy-brief-1 | Publication series | EPINOVA Policy Brief Series |
| IsSupplementedBy | https://github.com/EPINOVALLC/EPINOVA-Research | Repository | Supplementary repository and structural archive |
| References | Wu, S. (2026). Beyond model capability: A system-level framework for AI power | Policy brief | Referenced for the system-level definition of AI power and the six-dimensional AI power framework |
| References | Bergman, M. K. (2001). The deep web: Surfacing hidden value | Journal article | Referenced for the public web and deep web analogy |
| References | NIST AI Risk Management Framework materials | Government framework | Referenced for AI risk management, system context, intended use, and governance framing |
| References | European Union Artificial Intelligence Act | Regulation | Referenced for risk-based AI governance and regulatory context |
References
- Anderljung, M., Barnhart, J., Korinek, A., Leung, J., O’Keefe, C., Whittlestone, J., Avin, S., Brundage, M., Bullock, J., Cass-Beggs, D., Chang, B., Collins, T., Fist, T., et al. (2023). Frontier AI regulation: Managing emerging risks to public safety. arXiv. https://arxiv.org/abs/2307.03718
- Bergman, M. K. (2001). The deep web: Surfacing hidden value. Journal of Electronic Publishing, 7(1). https://doi.org/10.3998/3336451.0007.104
- Bommasani, R., Hudson, D. A., Adeli, E., Altman, R., Arora, S., von Arx, S., Bernstein, M. S., Bohg, J., Bosselut, A., Brunskill, E., et al. (2021). On the opportunities and risks of foundation models. arXiv. https://arxiv.org/abs/2108.07258
- European Union. (2024). Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). Official Journal of the European Union. https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng
- Gupta, A., Maynard, S. B., & Ahmad, A. (2021). The dark web phenomenon: A review and research agenda. arXiv. https://arxiv.org/abs/2104.07138
- Kapoor, S., Stroebl, B., Siegel, Z. S., Nadgir, N., & Narayanan, A. (2024). AI agents that matter. arXiv. https://arxiv.org/abs/2407.01502
- National Institute of Standards and Technology. (2023). Artificial intelligence risk management framework (AI RMF 1.0) (NIST AI 100-1). U.S. Department of Commerce. https://doi.org/10.6028/NIST.AI.100-1
- National Institute of Standards and Technology. (2024). Artificial intelligence risk management framework: Generative artificial intelligence profile (NIST AI 600-1). U.S. Department of Commerce. https://doi.org/10.6028/NIST.AI.600-1
- OECD. (2019). Recommendation of the Council on Artificial Intelligence (OECD/LEGAL/0449). OECD Legal Instruments. https://legalinstruments.oecd.org/en/instruments/OECD-LEGAL-0449
- OECD. (2024). OECD updates AI Principles to stay abreast of rapid technological developments. https://www.oecd.org/en/about/news/press-releases/2024/05/oecd-updates-ai-principles-to-stay-abreast-of-rapid-technological-developments.html
- Office of Management and Budget. (2025). Accelerating federal use of AI through innovation, governance, and public trust (Memorandum M-25-21). Executive Office of the President. https://www.whitehouse.gov/wp-content/uploads/2025/02/M-25-21-Accelerating-Federal-Use-of-AI-through-Innovation-Governance-and-Public-Trust.pdf
- Shevlane, T., Farquhar, S., Garfinkel, B., Phuong, M., Whittlestone, J., Leung, J., Kokotajlo, D., Marchal, N., Anderljung, M., Kolt, N., Ho, L., Siddarth, D., et al. (2023). Model evaluation for extreme risks. arXiv. https://arxiv.org/abs/2305.15324
- Wu, S. (2026). Beyond model capability: A system-level framework for AI power (Policy Brief No. EPINOVA–2026–PB–58). Global AI Governance and Policy Research Center, EPINOVA LLC. https://doi.org/10.67037/epinova.pb.2026.058