Particle.news

Microsoft Chief Says Anthropic's Fable Limits 'Don't Make Sense'

Nadella's July 16 rebuke could prompt customers to seek open or self-hosted AI.

Overview

  • On July 16 Microsoft CEO Satya Nadella told engineers that Anthropic's restrictions on requests to its Claude Fable model “don’t make sense,” marking a rare public critique of a major partner.
  • Anthropic restored global access to Claude Fable 5 on July 1 after U.S. export controls forced a temporary cut-off, but it kept Mythos 5 limited to vetted U.S. organizations.
  • To meet export and safety concerns Anthropic deployed a classifier that it says blocks reported jailbreak techniques and routes flagged queries to an older Opus model, a change that can cause benign requests to be blocked or answered by the older model.
  • Microsoft has deep commercial ties to Anthropic, including a $5 billion investment and large Azure commitments, while also promoting alternatives such as its Foundry catalog and in-house models to reduce customer dependence on a few frontier providers.
  • The dispute is pushing enterprises to weigh vendor lock-in, data rights, and model distillation rules and is driving interest in open-weight and self-hosted models as hedge options for customers and governments.