Ben Horowitz: The Fight Over Open Source AI
Ben Horowitz argues that open source AI is critical for safety, competition, and innovation, comparing it to successful open source technologies like Linux and the internet. He contends that banning open source would harm academia, stifle competition, and create dangerous monopolies, while open models allow distributed security improvements and prevent single-company control of AI.
Summary
Ben Horowitz discusses why open source AI represents the most important policy issue in AI development. He argues that historically, open source versions of technologies have been safer than proprietary alternatives—citing Linux and the internet as far superior to Windows in security. The key safety argument is that open source allows the entire community to work on problems like reward hacking, whereas proprietary labs focus on benchmarks and monetization rather than safety. Horowitz emphasizes that banning open source is impossible (it's just files on the internet) and would only prevent legitimate actors from using it while bad actors continue, similar to failed attempts to ban cryptography.
On competitive dynamics, he warns that companies like Anthropic are employing a "Microsoft playbook" where they enter application categories that succeed on their platforms, then undercut competitors. Open source models protect application developers from this predatory behavior. He notes that despite China's dominance in open source AI development, the U.S. can build competitive alternatives through companies like Thinking Machines and academia, but only if open source remains viable.
Regarding national security concerns, Horowitz argues that concerns about Chinese open source models are overstated compared to the risk of AI monopolies. He counters that having the last codebase is valuable even if China stops releasing new models, and that the real national security risk is concentration of AI power in a single company. He criticizes Anthropic's stated indifference to U.S. government concerns as a concerning display of monopolistic power.
On distillation and business models, he agrees with Aaron Levy that distillation is legitimate and beneficial, noting that outputs of AI models don't fall under copyright law. He suggests sustainable business models for open source labs include B2B enterprise services with smaller, specialized models and consumer applications tailored to specific use cases—more like Palantir than Anthropic.
Horowitz notes the AI market is less than 3% penetrated, so current competition between models is not existential—all players can grow simultaneously. He identifies customer support, specialized applications, and software development tools as areas where cheap open models create significant value. On AI art, he expresses optimism about democratization of creativity, drawing parallels to how new technologies (saxophone, electric guitar, drum machine) created new art forms rather than destroying existing ones.
About this episode
Ben Horowitz joins Theo Jaffee and Sofia Puccini to discuss one of the biggest debates in AI today: the future of open-source models. They examine the growing push to restrict open models, why Ben believes open source is critical for security, innovation, and competition, and what happens if a handful of frontier labs come to dominate the AI ecosystem. They also discuss distillation, AI monopolies, China's role in open-source AI, robotics, manufacturing, the economics of frontier models, and why Ben believes the AI market is still in its earliest stages. The conversation closes with a look at AI-generated art, creativity, and whether new tools will spark the next cultural renaissance.
Key Insights
- Horowitz argues that open source has historically been safer than proprietary versions (Linux vs. Windows) because entire communities can address security problems, whereas monopoly companies face no incentive to fix bugs that cannot be worked around.
- He claims that banning open source is technically impossible because it consists of files on the internet, and attempted bans would only prevent legitimate actors from using it while bad actors continue, similar to failed cryptography restrictions.
- Horowitz contends that Anthropic and other proprietary labs are using 'anti-competitive stances guised in a safety cloak' to push for open source restrictions that would primarily benefit their business models.
- He argues that application developers face existential risk if they build on proprietary models, as companies like Anthropic will enter their category, undercut them with subsidized versions, and either cut them off or overcharge them—making open source essential for ecosystem health.
- Horowitz states that Chinese open source models dominating the market is a natural consequence of American labs' proprietary business model focus, since the market leader never goes open source but number two eventually does.
- He claims the AI market is less than 3% penetrated, meaning all current players—Anthropic, OpenAI, open source models—can grow simultaneously, so competitive threats between them are currently minimal.
- Horowitz argues that the real national security risk is AI monopoly concentration in a single company rather than reliance on Chinese open source models, citing Anthropic's stated indifference to U.S. government concerns as evidence of dangerous concentrated power.
- He contends that open source model outputs create specific value for B2B enterprise applications where smaller, specialized models often outperform large proprietary models while being cheaper and faster.
Topics
Transcript
If you look at the history of the industry, the open source version of everything has been much safer. So the Internet and Linux were far safer than Windows, like by a lot. The safest thing for the world is that there's not one AI to rule them all, that there's AI for everybody. I think that open source is basically the best path towards that goal. It's probably the single most important policy issue around AI. It feels like we're at a stable equilibrium in the market, but the truth is the AI market is probably less than 3% penetrated. Should AI be open? Theo Jaffe and Sofia Puccini are joined by Ben Horowitz to discuss the battleowitz, who…
Full transcript available for MurmurCast members
Sign Up to AccessMore from The a16z Show
Sriram Krishnan on Open Source AI's Biggest Week Yet
Sriram Krishnan, former White House AI policy advisor, discusses how recent open-source AI models like Kimi K3 are reshaping the industry by increasing competition, creating pricing pressure on frontier labs, and raising important questions about security, distillation, and America's competitive position in the global AI race.
Building the Physical AI Stack | Travis Kalanick on TBPN
Travis Kalanick discusses his new company Atoms, which is building industrial AI solutions to automate mining, food production, and transportation. He raised $1.7 billion and explains how autonomous systems are increasing productivity and safety while reducing operational costs across these industries.
Travis Kalanick Is Back | Building the Future of Industrial AI
Travis Kalanick discusses his return to entrepreneurship with Atoms, a conglomerate applying industrial AI to transform manufacturing, logistics, and autonomous systems across food, mining, and transport industries. He reflects on lessons from Uber's 2017 crisis, his stealth-mode building strategy, and the framework of 'atoms-based computation' as the next frontier beyond software.
Why Physical AI Is the Next Frontier | Applied Intuition
Applied Intuition co-founders Kasser Younis and Peter Ludwig discuss how physical AI—intelligence deployed on machines that move—represents the next frontier of artificial intelligence, with applications across autonomous vehicles, mining, agriculture, and defense. They introduce Dana, their new platform designed to democratize autonomous systems development, and explore why physical AI presents fundamentally different engineering challenges than digital AI.
Hugging Face's CEO on Open Source AI, Model Routing, and the Future of Competition
Hugging Face CEO Clement Delange discusses open source AI's advantages over proprietary models, arguing it's inherently safer and better positioned for competition. He addresses government regulation of frontier models, the viability of open source business models ($100M ARR milestone), and envisions a future where AI routing across multiple specialized models replaces reliance on single frontier models.