Microsoft CEO Satya Nadella has publicly criticized Anthropic over the way its most advanced AI model, Claude Fable, handles user requests.
According to Firstpost, Nadella questioned Anthropic's decision to restrict certain requests, saying the limits "don't make sense." Reporting from MSN characterized his remarks more sharply, saying Nadella accused Anthropic of "editorially controlling" the model and argued that Claude Fable refuses too many requests and is overly restrictive. The criticism was also covered by Quartz and Yahoo Finance.
The comments land at a notable moment for the two companies. Anthropic is a major AI lab whose models are also offered through Microsoft's own products and cloud platform, making Nadella both a partner and, increasingly, a competitor. MSN reports that his criticism comes as Microsoft "doubles down on cheaper AI models," and Firstpost frames the remarks as arriving amid intensifying competition among AI developers.
At the heart of the dispute is a genuine industry tension. AI companies build guardrails so their models decline requests they judge harmful or inappropriate. But those same guardrails can frustrate users when a model refuses reasonable, benign tasks — a behavior often called "over-refusal." Nadella's argument, as reported, is that Anthropic has drawn those lines too tightly and is effectively imposing editorial judgments on what the model will and won't do.
The sources do not include a response from Anthropic, nor specific examples of the requests Nadella believes should be allowed.
Why it matters: how much AI models refuse — and who decides — is becoming a competitive battleground, and a top tech leader publicly challenging a rival's safety choices signals that "how restrictive is it?" may soon shape which AI tools businesses and consumers choose.