Nadella slams AI labs over distillation in 3-point critique

Microsoft Chief Executive Officer Satya Nadella criticized AI labs for complaining about model distillation, calling it hypocritical for companies that train on publicly available data to then restrict others from doing the same.

"While the great innovation that comes from model providers having fair use rights to train models on public data is needed, I find it ironic that the status quo is to then turn around and impose restrictive terms on distillation, and to reserve the right to learn from customer usage and interaction data," Nadella wrote in an X post on Sunday.

Model distillation involves training a smaller, less powerful AI model using the outputs of a more advanced one. Frontier AI labs including Anthropic, OpenAI and Google DeepMind have relied on publicly available writing, images and other data to train their flagship models — a practice that has drawn lawsuits from content creators and publishers. Nadella argued that if learning flows in only one direction, "owners of the learning infrastructure make all the money while creators of the knowledge get left out."

Though Nadella did not name any specific company, his comments appeared aimed at Anthropic. Earlier this year, Anthropic Chief Executive Officer Dario Amodei complained that Chinese model makers were using Claude to train their own models. Last month, the company wrote to US Senators Tim Scott and Elizabeth Warren alleging that Alibaba had carried out "the largest known distillation attack" on it to date. Anthropic said competitors can "acquire powerful capabilities from other labs in a fraction of the time, and at a fraction of the cost, that it would take to develop them independently."

The Data Ownership Debate

Nadella warned that companies relying on leading AI models are effectively handing over their proprietary data and then paying to use the same models. He said enterprises should own their AI infrastructure and institutional knowledge rather than depend on any single model vendor. Companies should conduct their own evaluations and maintain their own "learning loop" to improve AI capabilities continuously, he said.

"That is why enterprises need a real trust boundary for their human capital and token capital to compound," Nadella wrote. "And it is a hard boundary across which nothing crosses, not even the intelligence exhaust, without consent."

The debate highlights a growing tension in the AI industry over who controls the data and infrastructure that power modern models. Elon Musk has also criticized Anthropic's data practices, writing in a February X post that the company is "guilty of stealing training data at massive scale."

What It Means for Investors

For Microsoft, the comments reinforce the company's strategy of selling AI infrastructure through Azure rather than competing directly with frontier model makers. Microsoft has invested more than $13 billion in OpenAI but also offers access to models from multiple providers through its cloud platform. The tension between model providers and infrastructure owners could accelerate a shift toward enterprise-owned AI systems, reducing dependence on any single vendor. Companies building proprietary AI stacks — including Microsoft, Amazon and Google — stand to benefit as enterprises seek greater control over their data and model training pipelines.

This article is for informational purposes only and does not constitute investment advice.