The CEO of Microsoft, which has invested billions in Anthropic, criticized the company's new AI for censorship
Criticism from one of its largest investors is increasing pressure on Anthropic ahead of its IPO

In November 2025, Microsoft, under the leadership of Satya Nadella, invested $5 billion in Anthropic / Photo: FotoField / Shutterstock.com
“When you use the [AI model] Fable and it, for some random reason, refuses to fulfill a request, you can’t help but wonder: when was the last time a content-creation tool was subject to this level of editorial control? It doesn’t make sense,” said Microsoft CEO Satya Nadella. His remarks were reported by CNBC based on a transcript of a speech he gave to his staff. He was referring to Anthropic’s most powerful publicly available model.
What makes this criticism notable is that Microsoft itself is a major investor in Anthropic: late last year, the corporation invested $5 billion in the startup. Microsoft also has close ties to another AI developer—OpenAI. Around the same time, it valued its stake in OpenAI’s commercial entity at $135 billion. Meanwhile, Microsoft’s own AI assistant—Copilot—competes directly with both Fable and OpenAI’s ChatGPT.
AI without a lease
According to the head of Microsoft, customers should be able to adapt AI models to their specific needs at a reasonable cost and use their own data in a way that keeps it within the company.
“It can’t be that only two companies in the world own token capital (control the computing power needed to run AI models — Oninvest), while everyone else is forced to rent it,” Nadella said.
Through its Foundry cloud service, Microsoft provides access to thousands of AI models, including those developed by Anthropic and OpenAI. In addition, the company has begun merging the consumer and enterprise versions of Copilot into a single product. Nadella added that this should have been done “perhaps from day one.”
What's Happening with Stocks
Microsoft shares have fallen 17% since the beginning of 2026, while the Nasdaq Composite Index has risen 11% over the same period. In pre-market trading on July 17 in the U.S., the company’s stock is down more than 1%, pulling back after the previous day’s gains.
In early July, Wolfe Research analyst Alex Zukin lowered his annual price target for Microsoft from $570 to $525 per share, but maintained his “buy” rating on the stock. While Microsoft’s stock price used to depend mainly on the growth rate of its Azure cloud business, the latest quarter showed that investors are now demanding a more complete picture of the return on its record-high AI spending, Zukin said on CNBC.
The analyst emphasizes that Microsoft must clearly demonstrate exactly how it plans to monetize the integration of Copilot and other AI agents into its software products, and how these new technologies will accelerate revenue growth from software sales.
Context
Unlike its sister model, Mythos—which is available only to a limited group of Anthropic partners—Fable is equipped with robust security algorithms. These algorithms block attempts to copy (distill) the model itself, as well as potentially dangerous queries.
Criticism from a major investor like Microsoft is increasing the pressure on Anthropic: the startup needs to maintain its restrictions without letting frequent outages become a source of frustration for customers, according to Briefs Finance.
JPMorgan Chase CEO Jamie Dimon this week called the emergence of Mythos a “real problem” that Washington has so far managed to keep under control. According to Dimon, the deployment of the model requires government regulation, since widespread access to Mythos would be tantamount to freely distributing ballistic missiles.
This article was AI-translated and verified by a human editor




