Microsoft’s Nadella questions Anthropic Fable’s AI guardrails as model rivalry heats up

NEW DELHI: Sometimes the biggest AI debate is not about intelligence, but about who gets the last word. That tension came into focus after Microsoft chief executive officer Satya Nadella criticised restrictions built into Anthropic’s flagship Fable model, arguing that generative AI tools should not be “editorially controlled” when responding to users.

Speaking during an internal meeting with engineers working on Microsoft’s Copilot AI software, Nadella questioned why Fable refuses certain user requests. “If you use Fable, when it refuses for any random thing, it just is like, when was the last time you had a creation tool that was so editorially controlled? It doesn’t make sense,” he said, according to remarks obtained by CNBC.

The comments add a fresh dimension to the increasingly competitive AI landscape, where technology companies are balancing model safety with usability. Anthropic has introduced additional safeguards around Fable after tightening access earlier this year to comply with US export control requirements. The company has acknowledged that the updated protections may also block some harmless requests while aiming to reduce false positives.

The remarks also reflect a broader shift in enterprise AI strategies. Rather than relying solely on the most advanced frontier models, businesses are increasingly exploring lower-cost alternatives that can handle software development and enterprise workflows more efficiently. Chinese startup Moonshot AI added to that momentum this week by unveiling an open-source model it claims outperforms recent offerings from Anthropic and OpenAI.

Nadella’s criticism is notable given Microsoft’s commercial ties with Anthropic. The software giant invested $5 billion in the AI startup in 2025, while Anthropic committed to spending $30 billion on Microsoft’s Azure cloud platform. Microsoft also integrates Anthropic’s technology into parts of its Copilot ecosystem.

The Microsoft chief has repeatedly argued that enterprises should be able to build custom AI models using their own data without depending entirely on a handful of frontier AI companies. Microsoft has expanded its Foundry platform, which now offers developers access to more than 11,000 AI models, including those from Anthropic and OpenAI.

“It can’t be that there are only two companies in the world with token capital, and everybody else is renting it. It makes no economic sense,” Nadella told engineers, referring to the computing resources needed to run AI models.

He also acknowledged that Microsoft’s decision to combine its consumer and enterprise Copilot products should have happened much earlier. The company merged leadership for both businesses in March as it seeks to strengthen its AI strategy across personal and workplace applications.

The latest comments underline how competition among AI providers is evolving beyond raw model performance. As companies invest billions in artificial intelligence, debates over openness, safety controls, cost efficiency and customer choice are becoming just as important as the technology itself.