Nadella calls out AI labs like OpenAI and Anthropic for banning distillation while training on everyone else's data
Quick Answer
Microsoft CEO Satya Nadella criticizes AI labs like OpenAI and Anthropic for banning distillation while training on public data, calling it ironic.
Quick Take
Microsoft CEO Satya Nadella criticizes AI labs like OpenAI and Anthropic for banning distillation while training on public data, calling it ironic. He argues that this practice leads to economic value concentrating with infrastructure providers rather than knowledge creators, dubbing it the 'reverse information paradox.' Companies effectively pay twice for AI services, both financially and through valuable user interaction data.
Key Points
- Nadella calls distillation bans ironic as AI labs use public data under fair use.
- He highlights the economic imbalance favoring infrastructure operators over knowledge creators.
- Nadella refers to the situation as the 'reverse information paradox.'
- Companies pay for AI twice: financially and through user interaction data.
- Microsoft offers infrastructure for companies wanting to control their AI learning loops.
📖 Reader Mode
~1 min read
Jul 13, 2026
Microsoft CEO Satya Nadella calls out AI labs for banning distillation. Distillation is the process where smaller models learn from the outputs of larger ones. Providers like OpenAI and Anthropic ban the practice in their terms of service, targeting mainly Chinese AI companies. In a post on his blog, Microsoft CEO Satya Nadella now calls it "ironic" that these same providers train on public data supposedly under fair use, ban distillation, and learn from customer interactions. The result, Nadella says, is that economic value concentrates with the infrastructure operators instead of the companies that actually generate the knowledge.
In consuming intelligence, you are creating intelligence.
Satya Nadella
Nadella calls this the "reverse information paradox." Companies pay for AI twice. First with money, then with what he calls the "exhaust," meaning the corrections, ratings, and usage data from interactions with AI systems that reveal internal company knowledge. AI providers can learn from all of this and potentially build competing products. Nadella, of course, happens to have just the infrastructure to offer companies that want to control their own learning loop.
— Originally published at the-decoder.com
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from The Decoder
See more →
An AI model programmed nonstop for 19 days on a single MirrorCode task that cost $2,600 to run
Epoch AI's MirrorCode benchmark reveals Claude Opus 4.7 as the leader with a 56% solve rate, reconstructing a 16,000-line toolkit in 14 hours. Despite this, all models tested struggle with the most complex tasks, highlighting limitations in current AI capabilities. The single task consumed $2,600 over 19 days, raising questions about cost-effectiveness in AI development.
