The Little Tech Agenda for AI
Matt Perault and Colin McCune from a16z discuss the "Little Tech Agenda," advocating for AI policies that support startups. They argue for regulating the harmful use of AI, not its development, to foster competition and innovation while ensuring the U.S. maintains its global leadership against rivals like China.
The “Little Tech Agenda”: Advocating for Startups
The core of the discussion revolves around the “Little Tech Agenda,” an initiative born from the realization that while large, established technology companies have long had a voice in Washington D.C., startups and smaller builders were not being represented. The policy needs of a five-person startup are fundamentally different from those of a trillion-dollar corporation with a thousand-person compliance team.
Policy conversations often overlook the practical challenges faced by startups, which lack dedicated general counsels or policy heads. These companies face daunting technical and financial hurdles—securing data, compute, and talent—to compete with incumbents like Microsoft, OpenAI, or Google. The agenda’s goal is to champion regulatory frameworks that foster a competitive environment for these new entrants, rather than erecting barriers that entrench the market leaders.
Core Philosophy: Regulate Harmful Use, Not Development
A central tenet of their proposed framework is to regulate the harmful use of AI, not the development of the technology itself. This distinction is frequently misinterpreted as a call for no regulation. In reality, it advocates for leveraging and reinforcing existing legal structures.
“I actually can’t think of a single example across the portfolio in which we are arguing for zero regulation… Regulate use, do not regulate development somehow is interpreted as do not regulate.”
This approach suggests that when AI is used to violate existing laws—such as consumer protection statutes, civil rights laws, or criminal laws—those actions should be prosecuted. The focus is on the application of the technology, not on creating a preemptive, permission-based system for building models. This framework addresses most concrete fears about AI, as they typically involve actions that are already illegal. The argument is that this provides a robust, effective way to protect the public without stifling innovation at the source.
The Evolution of the AI Policy Debate
The current AI policy landscape was significantly shaped by a series of Senate hearings in the fall of 2023. Testimony from major AI CEOs, which included warnings about existential risks, “spooked Capitol Hill” and accelerated a push for restrictive regulation. This narrative was amplified by a decade of groundwork from groups like the effective altruist community, which fostered a climate of fear around the technology.
This led to proposals for unprecedented regulatory regimes for software, including:
- Licensing: Requiring companies to obtain a government license to build frontier AI models.
- Nuclear-Style Regulation: Treating AI development with the same level of oversight as nuclear energy, including international-style governance bodies.
- Bans on Open Source: Prohibiting or severely restricting the development and distribution of open-source models, driven by national security concerns.
These ideas were often promoted by large incumbents who negotiated voluntary commitments with the White House, leaving startups and future innovators without a seat at the table. This approach would have created a market with only a few dominant players, stifling the competitive ecosystem that venture investment aims to create.
Federal vs. State Roles: A Constitutional Approach
A clear division of regulatory labor between the federal government and the states is necessary to avoid a chaotic “50-state patchwork” that would be impossible for startups to navigate. The speakers propose a framework based on constitutional principles:
- Federal Government: Citing the Commerce Clause, they argue the federal government should oversee the development of AI and regulate the national market. This creates a single, coherent standard for building and deploying models across state lines.
- State Governments: States have a crucial role in policing harmful conduct within their borders. Most criminal law is enacted and enforced at the state level, making states the appropriate venue for prosecuting the misuse of AI for criminal acts.
The recent AI law passed in Colorado serves as a case study. It created a complex, risk-based assessment framework that the state’s own governor and attorney general now believe could be problematic. An alternative, more direct approach would be to simply codify that using AI to violate existing anti-discrimination law is illegal—a clear enforcement of “harmful use” rather than a bureaucratic process for developers.
National Security and Open Source
The debate around AI policy is inextricably linked to global competition, particularly with China. Restrictive policies, such as a licensing regime analogous to nuclear energy regulation (which has resulted in only a handful of new power plants in 50 years), would effectively cede the technological high ground to geopolitical rivals.
“If we do the same thing to AI… you don’t have the medical advancements, you don’t have the breakthroughs… Beyond that, we lose to China. Full stop.”
Open source is a key strategic asset in this competition. While initial concerns focused on preventing China from accessing US technology, the reality is that Chinese companies are developing powerful models independently. The strategic advantage lies in making American open-source models the global standard. This serves as a form of soft power and ensures the world is building on a platform aligned with US values, rather than creating a vacuum for Chinese products to dominate international markets.
The Path Forward
The policy conversation has shown positive evolution. The National AI Action Plan, for example, marks a rhetorical shift from a primary focus on safety to a balanced approach of ensuring the U.S. can “win while keeping people safe.” There is now broader, bipartisan support for open source and for creating federal resources (e.g., for compute and data access) to lower barriers to entry for startups.
The key priorities for the near future include:
- Establishing Federal Preemption: Creating a federal standard for model development to prevent a fragmented state-by-state regulatory landscape.
- Building Coalitions: Organizing the broader ecosystem of startups, investors, and other stakeholders to advocate for a pro-innovation agenda.
- Focusing on Affirmative Policies: Moving beyond reacting to restrictive proposals and proactively advancing ideas around workforce retraining, AI literacy, and increasing capacity for enforcement agencies to prosecute harmful use.