Apple is rethinking its approach to AI for Siri by exploring models from Anthropic and OpenAI. The tech giant, long known for developing its own AI, appears ready to make a strategic pivot. It seems that internal AI projects haven’t been progressing as quickly as hoped, and now the focus shifts to customised offerings built on external models.
Negotiations are underway to develop versions of Anthropic’s Claude and OpenAI’s ChatGPT that run on Apple’s servers. While the company also tried out Google’s Gemini, Claude stood out as the best performer. However, due to its steep price tag, Apple might lean towards OpenAI instead.
This shift has stirred up some internal concern. Some members of Apple’s foundation model team see the move as a vote of no confidence in their work, prompting thoughts of seeking opportunities elsewhere—in organisations like Meta or even OpenAI. High‑profile departures, including Tom Gunter’s exit and threats from the MLX team managing Apple’s AI framework, suggest that change isn’t coming without a bit of internal friction.
Should Apple make the switch, it plans to run these external models on its in‑house M‑series chip servers, ensuring both privacy and control. Smaller AI features, however, will remain on devices, preserving local processing. Beyond Siri, other AI projects are being scaled back. For instance, the Swift Assist coding model has been shelved, with future Xcode versions potentially incorporating ChatGPT or Claude. Additionally, John Giannandrea’s role has been narrowed, particularly concerning robotics and Core ML teams.
Even an acquisition talks for the AI startup Perplexity made the rounds, although no deal has been finalised. This reshuffle marks not only a shift in technology but also signals internal reorganisation aimed at keeping Apple at the cutting edge of voice and AI technology.
