
Satya Nadella told employees that Anthropic's Fable restrictions 'don't make sense,' as he pushes for cost-efficient models and internal control over AI development.
Satya Nadella told Microsoft engineers Wednesday that Anthropic's restrictions on its high-end Fable AI model make no sense, marking rare public criticism of a company Microsoft has invested $5 billion in.
"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?" Nadella said, according to a copy of the remarks provided to CNBC. "It doesn't make sense."
Anthropic's Fable model sometimes returns responses from an older version when users ask about certain topics like large-scale model development, the company's support page shows. Social media users have flagged rejections. Anthropic said in early June it was trying to cut false positives on blocked requests. Three days after launching Fable 5, it restricted access to comply with a U.S. export control directive and restored it on July 1, saying the new safeguards would flag slightly more harmless requests than before.
Microsoft declined to comment. An Anthropic spokesperson did not respond to a request for comment.
The remarks come as executives at Microsoft look for cost-efficient models that can handle software development inside the company without coming from the most well-funded labs. On Thursday, Chinese startup Moonshot AI released an open-source model it claimed beats recent Anthropic and OpenAI releases.
Nadella has argued that companies should be able to develop custom models cost-efficiently and use internal data without letting it flow to model-building firms. In a Sunday blog post, he cited Palantir CEO Alex Karp, who said on CNBC that technical organizations want to own the means of production.
"It can't be that there are only two companies in the world with token capital, and everybody else is renting it," Nadella told the engineers. "It makes no economic sense."
Microsoft's link to Anthropic is deep but not exclusive. In November, Microsoft said it was pumping $5 billion into Anthropic, and the startup agreed to spend $30 billion on Microsoft's Azure cloud. This year Microsoft launched Copilot Cowork, a business productivity assistant that uses Anthropic's models. Microsoft also offers Foundry, a service with more than 11,000 models including Anthropic and OpenAI.
Ties to OpenAI have frayed. The two companies drifted apart after the abrupt ousting and reinstatement of OpenAI CEO Sam Altman in 2023, with little notice to Nadella. OpenAI said in April it would bring its models to Amazon Web Services, beyond Azure, and Microsoft announced its own in-house coding model in June. Microsoft's stake in OpenAI's for-profit business was valued at $135 billion as of October.
Nadella also said merging consumer and corporate Copilot products was a good move, something "we should have done maybe day one." In March, he brought in former Snap executive Jacob Andreou to lead Copilot across both categories. In April Microsoft reported over 20 million paid seats for the work-centric Copilot, or 4% of its cloud-based Office customer base.
Drafted by a large language model from the source reporting linked above, then screened by automated publishing checks. It is not read by a journalist before publication. Some articles cite our Alpha Score. Verify prices and figures against the original source. Educational coverage, not personalized advice.