Microsoft’s CEO is pushing for a new kind of patent to address fears that leading AI models could become Trojan horses that steal intellectual property from corporate customers.
In a Sunday X post, Satya Nadella questioned how firms should protect their core IP at a time when top AI companies are hungry for more data to improve their large language models.
To keep their commercial customers at ease, the top AI providers, such as OpenAI and Anthropic, have strict policies to prevent training on copyrighted content. But earlier this month, Palantir CEO Alex Karp alleged that a growing number of businesses are worried that using top AI models could expose their IP to theft, all while paying hefty usage fees.
"I'm going to get no value, and they’re going to get my IP,” Karp told CNBC. Businesses "want to know they own the means of production, [and that] it’s not being transferred to someone else."
Karp’s comments sparked tech investor David Sacks, a White House advisor on AI, to point to the risk of an AI provider creating copycat products using data from their enterprise clients.
"Don’t think that can happen? Just look at Figma. According to The Information, Anthropic ‘blindsided’ its then-business partner with the launch of Claude Design,” Sacks tweeted. "This isn’t an isolated example. Anthropic has launched Claude Science, Claude Security, Claude Legal, and of course Claude Code — each expanding into categories previously served by companies building on top of their models."
The IP theft concerns could derail enterprise demand for the top AI models, the major cash cow for the leading providers, pushing companies to embrace open models. Surprisingly, the CEO of Microsoft —a major partner and investor in OpenAI—agrees with Karp.
“AI creates the reverse problem. In the AI age, the buyer risks giving away knowledge, just in order to use what they bought,” Nadella wrote. “You essentially pay for intelligence twice, once with money, and again with something even more valuable: the proprietary knowledge you must reveal to make that intelligence useful. The better you want the model to perform, the more of that knowledge you have to feed it!”
The essay naturally raises questions about whether the enterprise customer data flowing through AI models can truly remain private. Nadella suggests that data training is inevitable. "Models learn from 'exhaust,' the prompts people write, the tools agents use, and especially the corrections people make when the model is wrong. Every correction is distilled into institutional know-how," he said.
We reached out to Microsoft for comment. In the meantime, Nadella describes the problem as the “reverse information paradox,” where the user has to increasingly expose information to get the most value from the AI model — but the model itself remains opaque because it’s closed-source. In response, Nadella says we need an AI equivalent to patents that protect an inventor’s idea when it's disclosed to the public, preventing infringement.
Microsoft’s CEO also threw some subtle shade at top AI labs that train their models on data from the internet, sometimes without consent. “I find it ironic that the status quo is to then turn around and impose restrictive terms on distillation,” he said, referring to competitors using a large AI model to transfer the skills to a smaller one.
“If learning flows in only one direction, economic value converges toward the owners of the learning infrastructure rather than the creators of the knowledge itself. Therefore, it's imperative that we distribute the learning infrastructure to every firm so that they can control their own learning loop,” he added.