Podcast Summary

The AI Policy Podcast

Host: Gregory C. Allen (CSIS Wadhwani AI Centers)
Guests: Jai Ramaswamy (Chief Legal & Policy Officer, Andreessen Horowitz [A16Z]); Matt Perault (Head of AI Policy, A16Z)
Episode Theme: AI Regulation & Innovation—Venture Capital Perspectives and the Policy Needs of Startups
Date: March 3, 2026

Episode Overview

This episode dives into how venture capital firms, particularly Andreessen Horowitz (A16Z), approach AI policy, regulation, and support for startups (or "little tech") in the rapidly evolving field of artificial intelligence. Hosts and guests break down why “little tech" needs a distinct voice in policy debates, the philosophical and practical underpinnings of A16Z's regulatory positions, and the ongoing federal versus state battle over AI regulation. The discussion also includes A16Z’s role in supporting national security innovation and the risks and responsibilities tied to AI’s development and use.

Key Discussion Points & Insights

1. Guest Backgrounds and A16Z Firm Culture

Timestamps: 01:09–12:26

• Jai’s Journey:

  • Started with academic ambitions, pursued philosophy and history (01:09)
  • Pivoted to law, became a federal prosecutor in SDNY, focusing on white-collar crime, cybercrime, and illicit finance (02:46)
  • Moved to DC for policy work, including work on the Cybercrime Convention (03:12)
  • Transitioned to the private sector: led compliance at Bank of America and Capital One (04:08)
  • Shifted into tech, aligning with A16Z via blockchain/crypto interests (05:09)

• Matt’s Path:

  • Worked at Meta (Facebook), then led a tech policy center at Duke and UNC (10:49)
  • Notes A16Z's firm-wide ethos of continuous learning, “books are central to the firm” (09:46)
  • Points out parity in intellectual rigor between big tech firms and academia (10:52)

“When you listen to our cofounders...they’re often referring to books that they've read, to how they got up to speed on a new topic. So the learning process is really critical...”—Matt Perault (09:46)

2. Venture Capital 101: The A16Z Model & Importance of Policy

Timestamps: 12:26–19:08

• Venture capital (VC) is defined as a talent and technology business, not just a financial discipline (12:26)
• A16Z invests in risky, early-stage startups with the potential for outsized returns (14:05)
• Emphasizes “get rich slow”; most portfolio companies fail, a few generate enormous returns (15:00)
• Policy’s increasing importance: Post-2010, tech firms can't ignore politics; policy is “the long pole in the tent” (08:17)
• VC firms want to help startups build their own governmental and policy capacity, not act as outsourcers (07:03)

“The biggest risk we have as venture capitalists is missing out on a deal...you don’t win by just being safe.”—Jai Ramaswamy (15:00)

3. A16Z and National Security: The American Dynamism Portfolio

Timestamps: 19:08–24:00

• “American Dynamism” invests in companies focused on areas traditionally reserved for the government: defense, space, satellites, law enforcement, education, supply chain (21:12)
• Recalls the post-WWII government–tech relationship; A16Z wants to rekindle decentralized, fast innovation for national interests (23:20)
• Role of national security in A16Z’s mission: Not just defense, but broad societal infrastructure (20:33)
• Policy focus at A16Z began with crypto but expanded to broader tech policy as government skepticism increased (24:00)

“In more recent years, that skepticism [toward technology] has informed a lot of the policy debate... our concern was that we have this golden goose.”—Jai (24:00)

4. ‘Little Tech’ vs Big Tech: A16Z’s Regulatory Philosophy

Timestamps: 27:49–33:20

• “Little tech” defined: garage startups, pre-product, pre-funding innovators—A16Z wants these voices at the policy table (28:26)
• VC-fueled AI startups are sometimes “big” financially, but not institutionally or politically (31:16)
• Policy needs: Startups require certainty and clear, stable regulations—not vague, volatile, or enforcement-first regimes (30:00)
• A16Z often acts as a surrogate trade association, aggregating little tech perspectives for policy discussions, but still encourages portfolio companies to develop their own policy capacity (32:41, 34:52)

“If you’re a startup...your number one concern is about survival, and how do we provide a policy environment where they can flourish. You need certainty… and that’s super important.”—Jai (29:07)

5. Regulation: Federal vs State Roles & Federal Preemption

Timestamps: 36:46–47:15; 52:23–67:12

• A16Z opposes state-by-state development regulations, supporting federal preemption for a unified national AI market (37:15)
• Policy paper outlined a constitutional/federalist approach: Congress should regulate AI development and interstate commerce; states should target harmful use and criminal behavior within their jurisdictions (37:23, 39:40)
• Excessive compliance burdens from state-level patchwork could “crush” startups and hurt US competitiveness (39:40–40:07)
• Regulatory gaps should be assessed before new AI-specific laws are passed; many harms can be addressed with existing general laws like UDAP, civil rights statutes, or criminal law (43:08)

“States have an important role to play in regulating harmful use of AI… but if states are really active in regulating development, it’s hard to build a tool if you’re an AI developer…and have to build that tool for 50 different states.”—Matt (37:23)

• Dormant Commerce Clause limits how much states can regulate out-of-state activity; some state laws ignore these constitutional restrictions (54:19–58:35)

“You could have an open source developer in California...does that mean if it’s deployed in New York, the law applies there too? …States should not be in the business of creating a national standard.”—Matt (57:57–58:35)

6. Development vs. Use: Where Should Regulation Focus?

Timestamps: 41:13–48:39; 65:00–67:55

• A16Z argues legal focus should be on harmful use rather than preemptively restricting development
• Analogies: Regulate pilots not flying drunk (use), but airplanes themselves (development) only when the tech matures (41:13–41:41)
• Innovations can be stifled if regulation targets the “math” or early experimentation; wrong regulation at the “Wright brothers” moment of AI could kill the next big breakthrough (45:39–48:39)

“We’re at a very, very early stage…Had the federal government in the past tried to regulate the invention of flight, we wouldn’t even have Boeing.”—Jai (45:39–47:15)

• Some fields (bioweapons, mental health) may justify earlier intervention, but focus should often be on supply chain points (like DNA synthesis labs) or closing demonstrated regulatory gaps (51:39, 65:00)

7. Global Competition, Systemic Risks, and Open Source

Timestamps: 52:23–54:08

• Overregulation could spur other nations, especially China, to leap ahead—“open source hesitancy” is cited as a factor in China’s growing AI parity (52:25)
• Regulations should be humble and flexibility; be mindful of international dynamics and unanticipated consequences (53:08–54:08)

“We had a concern [about open source]…and now the concern is that we’re falling behind. These are changing, the technology is changing. All of this stuff is really, really early stage and…to make those determinations…is likely to be wrong. We just have to be humble about it.”—Jai (53:34)

8. Practical Regulatory Examples: Mental Health, Licensing, Worker Displacement

Timestamps: 63:25–66:34

• States have legitimate regulatory power in fields like licensing—mental health example: AI chatbots should not falsely claim licensure (63:48–65:48)
• Both state and federal government already have tools (like the FDA), but a gap analysis is key to new policies
• Worker displacement is an area of bipartisan consensus where sensible state-level action could be impactful (63:10)

9. Closing Reflections: AI’s Place in Society and Policy

Timestamps: 67:55–69:32

• AI is now the “control layer of computing”—it will mediate reality and deserves special attention for domestic innovation and national security (68:24)
• The US won’t get everything right, but a competitive, dynamic system is preferred to early, overreaching regulation
• The AI policy debate should be rigorous, intellectually honest, pragmatic, and forward-looking rather than purely reactive

“AI...will be the new control layer of computing...it will be something that mediates reality. We should care a lot that that system is produced here and is produced with all the checks and balances we have.”—Jai (68:24)

Notable Quotes & Memorable Moments

• On Tech & Policy

  • “You may not be interested in politics, but politics is interested in you.”—Jai quoting Trotsky (09:41)
  • “Venture capital, in some ways, is not really a financial discipline...it's a talent and technology business.”—Jai (12:26)
  • “If everybody thought it was a good idea, probably wouldn’t be a good candidate for your type of investment.”—Gregory (15:50)

• On Regulatory Balance

  • “Before you create something that's very specific to the particular AI technology, you should think about laws of general applicability.”—Jai (43:08)
  • “If you do the things people are pushing on the development side...the likelihood that you crush startups is extremely high.”—Matt (44:06)
  • “We’re humble enough to know we don’t have all the right answers, but...we want to put all our ideas out into the public sphere.”—Jai (67:12)

• On States’ Role & The Commerce Clause

  • “States aren't able to play the role of the federal government. There's a jurisdictional principle that gets at that: the dormant commerce clause...it's alive and well.”—Matt (54:19)
  • “In some ways, this AI debate is about a national market...there was a recognition that for the United States to exist and compete...there needed to be federal control over certain areas.”—Jai (60:54)

• On International Dynamics

  • “The danger is not likely to be what we think it is...and the technology is changing. So you monitor it and think about ways of balancing these issues.”—Jai (53:34)

Timestamps for Important Segments

• 01:09: Jai’s professional journey and early experiences
• 09:46: Matt on learning culture at A16Z and his path
• 12:26: VC model explained; the “get rich slow” approach
• 20:33: American Dynamism and national security investment
• 27:49: ‘Little tech’ defined; why startups need a policy voice
• 37:15: A16Z’s stance on federal vs state regulation
• 41:13: Regulating development versus use; analogy to aviation
• 52:23: Open source, innovation risks, and global competition
• 54:19: Dormant commerce clause and constitutional limits
• 63:25: State power in mental health and worker displacement
• 68:24: Closing advice—AI as the new "control layer" of computing

Overall Tone & Language

The conversation is open, intellectually curious, and emphasizes both humility and advocacy. Speakers often discuss the regulatory process as a balance of innovation incentives, national interest, competition, and societal safeguards, reflecting a rigorously analytical but pragmatic tone.

This summary captures the depth and detail of the episode, highlighting the central themes, core arguments, and memorable exchanges for listeners and policy professionals alike.