For all the heated debates about the potential decline of AI, there is one concern that is causing a lot of friction among AI enthusiasts in Silicon Valley. Their fear is that the big AI labs selling proprietary models somehow act as Trojan horses.
The concern is that, as startups and businesses deploy AI models from labs like OpenAI and Anthropic, the labs gain ever-increasing access to those companies’ most sensitive business information. Model makers can then use that information themselves, potentially becoming competitors to their clients. Those issuing such warnings range from VCs like Jason Calacanis to Palantir CEO Alex Karp.
Now, in a surprise blog post published on Sunday, Microsoft CEO Satya Nadella has joined the crowd. Nadella warns that AI users (“consumers” as he calls them) are paying twice. They deliberately spend to use AI tokens but also, inadvertently, transfer valuable data in the process.
“You’re basically paying for intelligence twice, once in money, and again for something more important: the proprietary information you have to reveal to make that intelligence useful. The better you want the model to perform, the more information you have to feed it!” he writes.
More dangerously, he says, businesses teach models about the nuances of their businesses.
He writes: “Models study the ‘exhaust,’ where people write, the tools agents use, especially the corrections people make when the model is wrong.”
This is “the kind of information a competitor would never buy,” but businesses are providing it.
Nadella says that if AI companies freely access the Internet to train their models, it’s fair for businesses to learn — or “distill” — those models in return. “Distillation” is the practice of using the output of a model to learn how it works and train a new, often cheaper, model based on that information. In February, Anthropic accused Chinese open source models of sending millions of notices to Claude as a way to improve their models, and urged the US government to ease export controls.
Nadella’s point is that model makers can’t have it both ways. It is hypocritical of them to train freely on real world data while restricting others from doing the same on their models.
“While innovation from model providers with the right to use it to train models on public data is needed, I find it strange that the current situation is to turn around and impose restrictive policies on distillation,” wrote Nadella.
Nadella is particularly concerned when model makers “reserve the right to learn from customer usage and interaction data.”
Nadella’s solution is the kind of thing a CEO of a major cloud provider would suggest. He wants companies to “retain ownership” of their data, including information, feedback, etc. So he encourages them to build themselves “proprietary learning environments” in the cloud (where their data is probably already stored anyway, and, ideally, it would mean Microsoft’s cloud, Azure). He’s also looking for companies to build on what he calls “orchestration layers” — essentially, a way to easily switch between AI models from different providers rather than being locked into one. Tools like AI “gateways” that allow companies to do this are becoming increasingly popular.
Although Nadella doesn’t use the words “open source” as a means of maintaining ownership, this is a clear subtext. However, there is another subtext.
Large corporations, many of which still have their own data centers in addition to using the cloud, are already moving to open source models installed on their premises (“on-prem,” in industry jargon). Idit Levine, founder and CEO of Solo.io – which makes networking and security software that helps businesses manage AI systems – says she can clearly see this change playing out with her clients. After exploring proprietary model builders, they started asking themselves: “Can I take an open source model and run it on-prem? It’ll do about 90% of what the big guy does. It’ll cost a lot less,” he tells TechCrunch. “They understand that, and they can control it.”
Solo.io technology was selected last year to be the technology powering the Linux Foundation’s Agentgateway project. His company counts businesses like T-Mobile, ADP, and SAP as clients. He sees companies increasingly incorporating open source models into the environment and sees it as the next big wave in the use of business AI.
He is not alone. Vercel (well-known as a platform for building and hosting websites, which recently added AI modeling tools) and OpenRouter (a company that helps developers of routing applications in all kinds of AI) are both seeing an increase in traffic to open source models. In fact, open models accounted for 29% of all traffic transferred to the Vercel portal last month.
With the CEO of Microsoft, the company that has invested in both OpenAI and Anthropic, now openly encouraging businesses to be wary of using proprietary models, we’ll bet this trend continues to grow. “In consuming intelligence, you create intelligence. And what you create must be yours,” Nadella wrote.
If you shop through links in our articles, we may earn a small commission. This does not affect our editorial independence.
