Satya Nadella has decided to break the corporate silence, exposing a major rift in the modern AI industry. The Microsoft CEO publicly accused OpenAI and Anthropic of blatant hypocrisy: while these labs "vacuum up" the entire internet under the banner of fair use, they are legally fortifying their systems against their own customers. The issue at hand is the ban on distillation—the process of training smaller, more efficient models using the outputs of their "older brothers." In Nadella's view, this isn't just a licensing dispute; it is a deliberate construction of digital feudalism.
The situation resembles a classic "information feedback paradox." Businesses pay for AI access twice: first with hard cash for tokens, and second by handing over their intellectual exhaust. Usage data, output ratings, and prompt refinements generated by companies during operation are absorbed into the providers' systems. While the labs use this feedback to improve their products, they slap the hands of customers who try to optimize their own costs by training local models on that same data.
Key points in the conflict of interest
AI labs prohibit using their APIs to train competing models. Customers generate valuable data that providers use for self-improvement. Barriers to distillation prevent companies from switching to cost-effective local solutions. Microsoft views this as a threat to tech democratization and an attempt to monopolize knowledge.
In the process of consuming intelligence, you inevitably create new intelligence, yet current rules ensure that all profits from this process stay in the pockets of infrastructure owners.
Microsoft’s stance in this debate is pragmatic. Nadella hints that the era of open-data romanticism is over, replaced by a fierce struggle for ownership of the insights a model generates. Labs continue to build fences around their APIs, despite having grown on publicly available content and continuing to benefit from user interactions. This is more than a legal conflict; it is a direct barrier for companies looking to reduce their reliance on cloud giants and cut costs by deploying compact, on-premise solutions.