What’s Fueling the AI Data Boom?
Inside this Issue
Behind the Billion-Dollar AI Data Headlines, Agent Security Gaps, and IA40 Recordings Now Available
AI data companies are reaching billion-dollar revenue run rates at remarkable speed. This week, Vivek Ramaswami and Sabrina Albert take a closer look at what’s driving that demand and what it will take to build lasting businesses.
We’re also sharing why Zico Kolter says smarter AI isn’t automatically safer, and former Anthropic lawyer Janel Thamkul’s advice on the risks founders accept in their contracts. Plus, all IA40 Summit session recordings are now available on YouTube.
Let’s get into it.
We’re also sharing why Zico Kolter says smarter AI isn’t automatically safer, and former Anthropic lawyer Janel Thamkul’s advice on the risks founders accept in their contracts. Plus, all IA40 Summit session recordings are now available on YouTube.
Let’s get into it.
Coral Garnick Ducken
Content Manager
Content Manager
Perspective
AI Data Companies Are On Fire
From zero to $1 billion in annualized revenue in under a year? Data for AI is producing some staggering growth numbers as model developers spend heavily on expert feedback and environments where agents can practice.
- Mercor passed a $2B run rate in June
- micro1 crossed $1B this month
- Proximal hit $200M annualized ~10 months after founding
But one strong month of sales isn’t recurring revenue, and a business supplying human experts has very different economics from one selling reusable software. This week, Vivek Ramaswami and Sabrina Albert break down what’s driving the demand for AI Data, what the headline numbers leave out, and what they're watching as the market develops.
IA40 Summit
AI Is Its Own Attack Surface: Zico Kolter on Securing Agents
What happens when your AI agent takes instructions from someone else? A malicious email or web page can turn a helpful assistant into an attacker with your credentials. And a more capable model won’t necessarily be harder to fool.
Madrona’s Vivek Ramaswami shares key takeaways from his IA40 Summit conversation with Gray Swan co-founder Zico Kolter, including what traditional security tests miss, why builders need to red team the model itself, and why waiting for a smarter model won’t solve the problem.
Gray Swan also just announced Arena 2.0, its global adversarial testing and red-teaming platform updated for the era of AI agents, which Fortune wrote about.
Madrona’s Vivek Ramaswami shares key takeaways from his IA40 Summit conversation with Gray Swan co-founder Zico Kolter, including what traditional security tests miss, why builders need to red team the model itself, and why waiting for a smarter model won’t solve the problem.
Gray Swan also just announced Arena 2.0, its global adversarial testing and red-teaming platform updated for the era of AI agents, which Fortune wrote about.
All the IA40 Summit session recordings
are now available on YouTube.
are now available on YouTube.
Founded & Funded
Anthropic’s Former Lawyer on AI’s Assumption of Risk Problem
When an AI agent goes off script, several parties could be responsible: the user, the founder’s company, the model provider, and any third-party system the agent touches. Founders are writing contracts and designing products around that open question.
At our 2026 CEO Summit, Madrona General Counsel Joanna Black asked Janel Thamkul, former deputy general counsel at Anthropic, where responsibility actually lands compared to where the average founder might assume. Her answer is that right now it is whatever you have negotiated, that agreements range from standard paper to highly negotiated, and that the industry has moved away from the early “use at your own risk” terms.
The same logic applies to the terms a founder writes for their own customers. Janel explains what those terms reveal about a company, including what it indemnifies and which risks it pushes onto the customer, and why copying another company’s terms can leave a founder accepting risks they don’t realize.
The full conversation is now available on Founded & Funded.
At our 2026 CEO Summit, Madrona General Counsel Joanna Black asked Janel Thamkul, former deputy general counsel at Anthropic, where responsibility actually lands compared to where the average founder might assume. Her answer is that right now it is whatever you have negotiated, that agreements range from standard paper to highly negotiated, and that the industry has moved away from the early “use at your own risk” terms.
The same logic applies to the terms a founder writes for their own customers. Janel explains what those terms reveal about a company, including what it indemnifies and which risks it pushes onto the customer, and why copying another company’s terms can leave a founder accepting risks they don’t realize.
The full conversation is now available on Founded & Funded.
Portfolio Openings
Madrona. 999 Third Avenue, Suite 3400, Seattle, WA 98104
This message was sent to you as a member of the Madrona Venture Group community. If you no longer wish to receive these types of messages from us in the future, then please feel free to opt-out by clicking .
