Apple throws in the towel on Siri AI to adopt private ChatGPT or Claude models

After a period of testing its homebrew Siri AI efforts against the likes of ChatGPT, Claude, or Google Gemini, Apple is reportedly ready to throw in the towel on its Foundation Models.

It is now in talks with OpenAI and Anthropic to adopt their large language models (LLM) to power the AI features it promised for Siri ever since it unveiled Apple Intelligence nine months ago.

Unfortunately, Apple has been unable to deliver the Siri AI upgrade with its in-house LLM, even on its most powerful handsets like the iPhone 16 Pro Max, and eventually put it off twice to 2026. Its AI team has apparently been either demoralized by the lack of clear directions, or enticed by the multimillion pay packages that companies like Meta or OpenAI promise its engineers to poach them.

When the mostly derivative Apple Intelligence features failed to spur an iPhone upgrade cycle and the homebrew Siri LLM got delayed, Apple replaced the head of its AI team and ordered a performance review. It turned out that the Siri AI features it promised would be served best by Anthropic's Claude LLM rather than Apple's Foundation Models.

Next in the performance line was ChatGPT, and this is why Apple is currently in talks with both Anthropic and OpenAI to hash out an adoption agreement to power Siri with their chatbot technologies.

Apple's main consideration when adopting a third-party service is, as usual, the privacy of iPhone owners. It has probed the idea of hosting Anthropic or OpenAI code on its own Private Cloud Compute server clusters and asked them to develop custom ChatGPT and Claude models that will leave it controlling the privacy settings of future Siri AI users.