Governing the AI Transition: Lessons from the 1996 Telecommunications Act
As artificial intelligence (AI) continues to reshape the landscape of technology and society, there are more than 300 bills related to AI currently introduced in the U.S. Congress, alongside approximately 1,200 in state legislatures. The challenge of legislating during a technological transition is both crucial and fraught with risks.
The Importance and Risks of Legislation
Legislation is vital for protecting the public interest, ensuring that rules and regulations guide corporate behavior rather than leaving companies to act solely in their own interest. However, there is a significant risk in that lawmakers often project the future based on current knowledge, which can stifle the necessary agility in a rapidly evolving technological environment.
The Telecommunications Act of 1996
The last major legislative effort during a technological transition was the Telecommunications Act of 1996, signed into law by President Clinton. This act updated the Communications Act of 1934 and serves as a critical case study in governing technology transitions. At that time, the transition from analog to digital technology was a destabilizing event, collapsing long-established business categories and restructuring markets.
Instead of attempting to predict future technological paths, the act focused on market structures that would guide those developments. Congress empowered the Federal Communications Commission (FCC) to identify and address chokepoints that could obstruct effective competition.
Lessons from the 1996 Act
One significant takeaway from the Telecommunications Act is that scale won. Over the years, market forces gravitated towards larger firms, undermining the act’s aspirations for competitive rivalry. This resulted in increased concentration in broadcasting and telephony, often at the expense of localism.
The FCC transformed from a public interest regulator overseeing monopolies to a referee promoting competition across converged markets. The act did not eliminate regulation; rather, it redirected regulatory priorities.
The Emergence of the Internet
Notably, the Telecommunications Act did not address the emerging internet, which thrived due to its inherent openness. This openness enabled companies like Google and Facebook to rise as competitors, leveraging open standards and network access to foster innovation. However, as these platforms grew, they constructed closed ecosystems that stifled the very competition that helped them succeed.
AI as a Continuation of Platform Business Models
The emergence of AI was not a sudden miracle but a natural progression within the online platform economy. AI’s business model is fundamentally about prediction, driving the need for advanced algorithms to optimize user engagement and behavior.
The Telecommunications Act highlighted the importance of controlling choke points, a lesson that resonates in today’s AI landscape. Whereas telecom bottlenecks were physical in nature, AI chokepoints are more economic and subtle, posing similar exclusionary threats.
The AI Stack and Its Implications
AI operates as a stack of interdependent layers—from microprocessors powering algorithms to cloud computational capabilities and specific applications. Each layer presents opportunities and potential choke points that can stifle innovation.
Commoditization and Economic Strategies
A significant shift is occurring in the AI landscape, with foundation models beginning to show signs of commoditization. As competition intensifies and open-source models gain traction, economic realities compel firms to seek higher-margin activities in AI applications.
This transition mirrors patterns from the Telecommunications Act era, where incumbents sought to enter new markets and bundle services to retain consumer loyalty. By embedding proprietary AI applications into workflows, dominant companies can create switching costs that enhance their market power.
Policy Recommendations for AI Governance
The lessons from the Telecommunications Act offer valuable insights for AI governance. A focus on anti-competitive control of essential capabilities is paramount. A two-step regulatory framework should be established, overseen by an independent expert agency:
- Ex ante non-discrimination: Ensure essential inputs are accessible on fair and transparent terms. This includes compute access, model licensing, and distribution neutrality.
- Ex post governance: Once openness is achieved, implement safety evaluations, consumer protections, and antitrust enforcement against exclusionary practices.
As the AI landscape evolves, it is crucial to recognize that technological transitions are not merely technical challenges but also involve power dynamics. If concentrated AI power remains unaddressed, the consequences extend beyond market structures to impact labor markets, national security, and the very fabric of democracy.
In conclusion, the Telecommunications Act of 1996 serves as a pivotal reference point. Addressing the concentration of power in AI is essential to fostering an environment where innovation can thrive, ultimately benefiting society as a whole.
