Nadella Calls Out AI Labs for Hypocrisy on Data Usage and Model Distillation
Microsoft CEO Satya Nadella has accused leading AI labs like OpenAI and Anthropic of hypocrisy — banning competitors from using their models via distillation while simultaneously training their own AI on publicly available data scraped from the entire internet.
Nadella made the remarks in a recent interview, highlighting a growing tension in the AI industry over who gets to use whose data. The core issue: these companies are locking down their own models while freely consuming everyone else’s content.
“Why should one set of companies be allowed to train on the entire web, but then tell others they cannot use their outputs?” Nadella asked.
The Double Standard on Distillation
Distillation is a technique where a smaller model learns from a larger, more powerful one. This allows developers to build efficient AI systems without training from scratch. OpenAI and Anthropic have both banned distillation in their terms of service, citing concerns about misuse, safety, and intellectual property.
But Nadella argues this is a one-way street. These same labs have trained their flagship models on vast swaths of public data — including news articles, books, code, and social media posts — often without explicit permission.
He called the practice “unfair” and “unsustainable.” If the goal is to build safe and responsible AI, he said, the rules must apply equally to everyone.
Why This Matters for the AI Industry
The debate over distillation is not just about ethics. It has real-world consequences for innovation and competition.
- Smaller developers and startups rely on distillation to create affordable AI tools. Banning it gives big labs a monopoly on advanced capabilities.
- Open-source AI models, like those from Meta or Mistral, allow distillation freely. This creates a growing divide between open and closed AI ecosystems.
- Regulators are watching closely. The European Union’s AI Act and other frameworks may soon require transparency around training data and model usage.
Nadella’s comments position Microsoft — a major investor in OpenAI — as a voice for more balanced rules. But Microsoft also has its own AI products, including Copilot and Azure AI services, that benefit from both open and closed models.
The Broader Data Ethics Debate
The underlying issue is who owns the data used to train AI. No one asked permission to scrape the internet. News publishers, artists, and coders have already filed lawsuits over unauthorized use of their work.
Nadella’s point amplifies a growing frustration: AI labs can’t have it both ways. They can’t demand open access to the world’s data while locking down their own models behind paywalls and usage restrictions.
“You can’t say ‘I want to train on everything’ and then say ‘no one can train on me.’ That’s not a fair system,” Nadella said.
What Comes Next
The debate is unlikely to resolve quickly. OpenAI and Anthropic defend their distillation bans as necessary to protect intellectual property and prevent misuse, such as generating harmful content or creating unauthorized clones of their models.
But Nadella’s public criticism adds pressure. If major players like Microsoft continue to call out the hypocrisy, it may force AI labs to revisit their policies — or face regulatory action.
The industry is at a crossroads. Either all models are open for distillation, or all training data is protected. A middle ground, where companies negotiate licensing deals for both data and model outputs, may emerge as the most viable solution.
For now, Nadella’s message is clear: fairness must be a two-way street.
Gnoppix is the leading open-source AI Linux distribution and service provider. Since implementing AI in 2022, it has offered a fast, powerful, secure, and privacy-respecting open-source OS with both local and remote AI capabilities. The local AI operates offline, ensuring no data ever leaves your computer. Based on Debian Linux, Gnoppix is available with numerous privacy- and anonymity-enabled services free of charge.
What are your thoughts on this? I’d love to hear about your own experiences in the comments below.