Apple’s smarter Siri with AI could be powered by OpenAI or Anthropic

Apple

Apple is in early talks with OpenAI and Anthropic to power a next-generation version of Siri, a move that could sideline its own AI ambitions — at least temporarily. According to Bloomberg’s Mark Gurman, the company is exploring whether to license customised large language models (LLMs) from the two AI firms, which would run on Apple’s private cloud infrastructure rather than external platforms like AWS or Azure.

By keeping LLMs on Apple silicon-powered servers, Apple aims to maintain tighter control over data privacy — a key marketing pillar for the brand. The revamped Siri effort is being spearheaded by Mike Rockwell and Craig Federighi, following CEO Tim Cook’s reported loss of confidence in AI head John Giannandrea.

While Apple is still developing its own in-house model — internally dubbed “LLM Siri” — the company has acknowledged that third-party tech, particularly Anthropic’s, is currently ahead. VP of corporate development Adrian Perica has been personally leading discussions with Anthropic, but sources say the two sides remain far apart on pricing, with the startup demanding a multi-billion dollar annual deal that escalates sharply over time.

OpenAI remains a potential fallback, especially as Apple continues to weigh the financial and strategic implications of outsourcing core AI functions.

The behind-the-scenes tension is starting to show. Morale within Apple’s Foundation Models team has reportedly dropped, with some employees feeling unfairly blamed for lagging progress in generative AI. Others have left, citing Apple’s comparatively conservative compensation. In one high-stakes moment, the entire team behind Apple’s open-source MLX framework nearly quit, before the company rushed in with counteroffers, as per the Bloomberg report.