The Politics of Fragmentation and Capture in AI Regulation
The political economy of artificial intelligence (AI) regulation is increasingly shaped by the strategic behaviors of various stakeholders, including governments, technology companies, and other influential agents. As AI systems, particularly those based on large language models (LLMs), evolve, international discussions surrounding regulations intensify.
Recent developments in AI regulation highlight a growing need for understanding how different national regulatory frameworks will interact. For instance, the European Union has passed the AI Act, while the United States has seen a series of executive orders and state-level proposals, alongside China‘s stringent national data controls. This landscape raises critical questions regarding the future of global AI governance amidst a fragmented regulatory environment.
The Local Game: Four Paths for Individual Jurisdictions
The regulation of AI can be understood through a local-level model that outlines the possible paths available to different jurisdictions:
1. No Local Regulation
Some jurisdictions may opt against regulating AI, perceiving the associated harms as minimal, experiencing regulatory capture, or finding enforcement costs prohibitively high. This laissez-faire approach allows companies to operate freely, potentially exposing citizens to unregulated risks.
2. Compliance and Local Adaptation
Proactive jurisdictions may establish enforceable regulations that companies comply with, leading to the ideal scenario where businesses adapt their operations to fit local legal frameworks. This scenario is more likely when the costs of evasion are greater than compliance.
3. Partial Evasion and Regulatory Gaps
In many cases, some companies comply with regulations while others evade them. This disparity arises from governments lacking the capacity or political will to enforce rules effectively, leading to uneven consumer protections and distorted market competition.
4. Market Withdrawal
When regulations become too burdensome, companies may opt to exit the market entirely, as seen when Google and Meta withdrew from China to avoid compliance with stringent laws, highlighting the potential downsides of strict regulatory environments.
The Global Game: Four Futures for AI Governance
Expanding the analysis to the global level reveals the complexities of international regulatory interactions:
1. Multiple Local Regimes
In this scenario, many countries establish their regulatory frameworks, permitting some level of arbitrage or evasion while maintaining political autonomy. This “benign fragmentation” respects national sovereignty and allows for diverse regulatory approaches, such as stricter consumer protection in the EU compared to the U.S.
2. International Harmonization
As regulatory divergences become increasingly pronounced, the need for international harmonization may arise. Governments may seek to bridge differences through treaties and coordinated rulemaking, aiming to reduce compliance burdens for companies operating across borders.
3. Unilateral Imposition (The “Brussels Effect”)
Occasionally, a powerful jurisdiction can set de facto global standards through strict early regulation. This phenomenon, known as the “Brussels Effect,” compels companies to adopt the highest standard worldwide, as seen with Apple’s shift to USB-C chargers following EU mandates.
4. Global Fragmentation (Splinternet of AI)
In a fractured scenario, countries may enforce fully sovereign AI regimes, leading to significant regulatory divergence. Companies may be forced to create separate products for different markets, which can stifle innovation and increase costs.
Real-World Outlook on Current Events: Strategic Fragmentation
Recent events, particularly Executive Order 14179 issued by the Trump administration in January 2025, exemplify strategic fragmentation dynamics. This order rescinds previous AI safety measures and mandates the development of an “AI Action Plan,” signaling a shift towards prioritizing local industry interests.
This regulatory posture aligns with the interests of the U.S. AI industry, potentially inviting companies to relocate operations to the U.S. for a more permissive environment. For countries pursuing stricter AI regulations, this poses significant challenges in maintaining their regulatory approaches.
As jurisdictions assert their regulatory independence, the resulting fragmentation may undermine the economies of scale necessary for efficient AI development, compelling companies to produce jurisdiction-specific models. In the long term, this may lead to selective harmonization among allied countries that seek to balance sovereignty with economic efficiency.
Ultimately, the likelihood of a globally harmonized AI governance regime remains low, given its entanglement with geopolitical competition and economic sovereignty. Instead, a world characterized by strategic fragmentation is anticipated, where jurisdictions prioritize their regulatory independence while selectively cooperating in areas of mutual benefit.
