a16z AI Policy Brief

AI benchmarks are becoming increasingly central to policy debates. But much of the value of these benchmarks depends on how they are built: what they measure, who develops and runs them, and how quickly they evolve alongside the models they evaluate.In this episode, Matt Perault speaks with Rayan Krishnan, cofounder and CEO of Vals, and Glenn Parham, head of public sector, about their work building AI evaluation tools and partnerships for industry and government. They discuss why static tests quickly become obsolete, why credible evaluations require deep domain expertise and independent testing, and how a growing market of evaluators is moving the field beyond “evaluations by vibes” toward more rigorous, data-driven evidence for policymakers.Topics covered:00:00: Intro02:00: What is Vals?05:00: Keeping benchmarks current08:00: New strides in AI evaluation11:00: Measuring AI’s cyber capabilities14:00: Why government needs independent evaluation21:00: The growing market for AI benchmarks26:00: What benchmarks can—and cannot—measure29:00: Building trust in AI benchmarks35:00: Why AI performance varies across domains and tasks37:00: Bringing AI evaluation to government41:00: From “evaluations by vibes” to better evidenceResources:Subscribe to the a16z AI Policy Brief: https://a16zpolicy.substack.com/Follow Matt Perault on X: https://x.com/MattPeraultFollow Rayan Krishnan on X: https://x.com/RayanKrishnanFollow Glenn Parham on X: https://x.com/glenn__parhamFollow Vals.ai on X: https://x.com/ValsAIThe content here is for informational purposes only, should not be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security, and is not directed at any investors or potential investors in any a16z fund. Please note that a16z and its affiliates may maintain investments in the companies discussed in this podcast. For more details, including a link to our investments, please see a16z.com/disclosures. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit a16zpolicy.substack.com

AI is reopening a core question from the development of the web: how to preserve the freedom to learn while giving publishers meaningful control over how their works are made available.In this episode, Matt Perault is joined by Derek Slater, cofounder of Proteus Strategies and an expert on information access, copyright, and free expression, to examine this question and explain why it matters for the future of AI.The web worked because people could access, read, analyze, and build on lawfully available information. Standards like robots.txt helped manage the balance between openness and control at scale, giving publishers a way to express preferences without requiring every builder to negotiate permission website by website.AI is now testing that equilibrium. Some proposals would restrict not only unlawful access, like when AI developers circumvent paywalls to get data, but also lawful learning from public information through expanded copyright theories, terms of service, technical barriers, or licensing requirements. Derek and Matt separate those issues, including the difference between training models on lawfully accessed data, producing infringing outputs, using AI tools to summarize content a user can already access, and breaking through access controls.For Little Tech, this question is fundamental. Access to data operates as a form of startup capital, allowing new companies to develop products and compete. But if AI companies can’t learn without negotiating expensive licenses or if large pools of data are entirely off limits to AI learning, then only the biggest, most-resourced companies will be able to survive.Topics covered:00:00: Introduction01:45: The freedom to learn and AI data access04:16: What the early web can teach us about openness, control, and contested norms08:48: How robots.txt helped publishers express preferences at scale11:43: Why voluntary standards worked for search engines and publishers13:50: How freedom to learn applies beyond technology and copyright debates16:02: The publisher POV on traffic, monetization, and value exchange18:36: Why AI agents are raising new questions about user control20:36: The difference between protecting publishers and limiting lawful AI-assisted reading21:42: How copyright law applies to AI training inputs and model outputs28:08: How contracts and terms of service can attempt to restrict lawful learning31:12: The limits of “learning” as a defense33:53: How licensing markets are evolving around data access and AI outputs37:12: Technical collaboration and the future of robots.txt-style standards for AI39:43: Why data access is a Little Tech issue42:25: Public policy guidance for preserving the freedom to learn45:29: Closing thoughtsDisclosure: Derek Slater has previously provided consulting support to Andreessen Horowitz. The views expressed here are his own, and this conversation was not part of a paid engagement.Resources:Subscribe to the a16z AI Policy Brief: https://a16zpolicy.substack.com/Follow Matt Perault on X: https://x.com/MattPeraultFollow Derek Slater on X: https://x.com/derekslaterThe content here is for informational purposes only, should not be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security, and is not directed at any investors or potential investors in any a16z fund. Please note that a16z and its affiliates may maintain investments in the companies discussed in this podcast. For more details, including a link to our investments, please see a16z.com/disclosures. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit a16zpolicy.substack.com

Marc Andreessen joins Navin Girishankar, president of the economic security and technology department at the Center for Strategic and International Studies (CSIS), for a wide-ranging conversation on artificial intelligence, productivity growth, national competitiveness, and America’s technological future.In their conversation, Marc argues that while AI has already begun reshaping the economy, the largest impacts are still ahead. He explores how AI could dramatically expand access to intelligence, improve productivity, and transform industries ranging from healthcare and education to law and software development. At the same time, he warns that many of the biggest barriers to progress are not technological but institutional, driven by policy choices and infrastructure constraints.The discussion also covers the global AI race, U.S.-China competition, export controls, energy, reindustrialization, and the role of government in fostering innovation. Along the way, Marc shares his views on technological progress and why he believes America still has an opportunity to lead the next wave of economic growth.Topics covered:00:00: Introduction: acceleration, transition, and policy00:02: The AI boom: optimism versus utopianism00:04: AI as an intelligence equalizer00:08: Education and institutional reform00:13: Productivity by sector: blue sectors and red sectors00:18: AI infrastructure constraints00:22: Tariffs and behind-the-border constraints00:26: Model export controls and the Mythos case00:33: The technological imperative and policy tradeoffs00:37: Lessons from Netscape and encryption export controls00:43: U.S.-China competition, open source AI, and civil-military fusion00:50: Public sector reform and government capability00:53: AI for public policy00:55: Industrial renaissance and American Dynamism1:00: Closing thoughtsResources:Subscribe to the a16z AI Policy Brief: https://a16zpolicy.substack.com/Follow Marc Andreessen on X: https://x.com/pmarcaFollow Navin Girishankar on X: https://x.com/ngirishankarFollow CSIS on X: https://x.com/CSISPlease note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see http://a16z.com/disclosures. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit a16zpolicy.substack.com

Policymakers often say they want to support startups. Republicans say it. And Democrats say it.But despite repeatedly saying they support startups, policymakers often propose rules that make it harder for those startups to build and compete.In this episode, Collin McCune, head of government affairs, joins Matt Perault, head of AI policy at a16z, to discuss the political economy of Little Tech: the structural dynamics that produce a policy process that so consistently rewards the companies with the most time, money, and access. In sum, it favors Big over Little.They also talk about the impact of the political economy of Little Tech on consumers. When regulation locks in incumbents, people end up with fewer choices, products that are lower quality and less innovative, and higher costs.Topics covered:01:00 The paradox of policymaker support for startups03:30 Why startups are underrepresented in the policy process05:40 How feedback on bills works in practice07:30 Why “industry” feedback often misses Little Tech09:45 The costs of showing up late to policy debates12:00 Audits, impact assessments, and invisible compliance costs15:30 How large policy teams create incumbent advantage18:30 Why audits can be harder than they look21:15 Policy ideas to better account for startup costs23:45 Why startup competition matters for everyday people25:45 Bright spots for Little Tech in Washington27:30 Lessons from Dodd-Frank and the risk of repeating them in techResources:Subscribe to the a16z AI Policy Brief: https://a16zpolicy.substack.com/Follow Matt Perault: https://x.com/MattPeraultFollow Collin McCune: https://x.com/Collin_McCunePlease note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see a16z.com/disclosures. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit a16zpolicy.substack.com

Policymakers have spent years talking about rebuilding America’s industrial base, reshoring critical supply chains, strengthening defense production, and reducing U.S. dependence on China. But recognizing the need to build is not the same as having the ability to do it.Erin Price-Wright, general partner on Andreessen Horowitz’s American Dynamism practice, joins the AI Policy Brief to make the case that AI isn’t just a software story. It’s the defining factor in the sectors that determine whether the U.S. can build, power, and defend itself in the decades ahead.She and Matt Perault discuss how AI can help make the math work for building in the U.S. again—from accelerating mine permitting and coordinating complex industrial projects to designing factories, lowering the cost of automation, and bringing robotics to more factory floors.They also discuss where policy needs to catch up: the laws and regulations that make it too hard and slow to build new factories in the U.S. and defense procurement that still favors incumbents over startups.Finally, they discuss how the debates over data centers and jobs will shape whether America’s reindustrialization effort succeeds.The takeaway: if the U.S. gets the policy environment right, AI can strengthen the industrial base, help create new kinds of jobs, and give America a powerful competitive advantage.Topics covered:00:00: Intro00:54: Erin’s work investing in AI for the physical world01:47: Why AI and reindustrialization are converging now03:07: Applying AI to mining and critical minerals08:28: What Ukraine reveals about defense production18:13: How startups are breaking into government markets20:16: Bringing a factory mindset to critical sectors and complex systems24:39: How robotics can expand factory automation30:22: What still makes it too hard to build in the U.S.32:49: Using AI to design better, cheaper manufactured goods35:05: Why data centers matter for reindustrialization39:46: How compute could help modernize the grid and lower costs for consumers42:50: Why AI could create new industrial jobsResources:Subscribe to the a16z AI Policy Brief: https://a16zpolicy.substack.com/Follow Matt Perault: https://x.com/MattPeraultFollow Erin Price-Wright: https://x.com/espricewrightPlease note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see a16z.com/disclosures. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit a16zpolicy.substack.com