Apple is reportedly considering using generative AI models from OpenAI and Anthropic to power an updated version of Siri, according to Bloomberg. People familiar with the discussions say the company is asking both firms to train custom versions of their language models — ChatGPT and Claude — that could run securely on Apple’s private cloud infrastructure.

However, if this plan moves forward, Siri could rely on generative AI models developed outside Apple. This could be a major departure from the company’s typical approach of building everything in-house.

Apple wants to run external models on its cloud servers

Bloomberg reports that Apple is asking for versions that can run on its Private Cloud Compute, a secure infrastructure built with Apple’s chips. These servers are designed to give Apple more control over the technology while protecting user data.

Apple believes this setup will help preserve privacy and maintain its usual level of control. The company has already tested whether these models can work inside its system.

The setup would mirror what Samsung is already doing. Samsung’s Galaxy AI uses a mix of its features and Google’s Gemini models. If Apple goes through with this plan, Siri would similarly become a blend of third-party intelligence and Apple-specific integration.

Apple hasn’t made a final call yet

No decision has been made yet. The company is still developing in-house models and planning a full Siri upgrade for 2026. Apple had initially aimed to roll out the new Siri by 2025, but delayed the launch.

At the same time, Apple is actively exploring options. In addition to its talks with OpenAI and Anthropic, Bloomberg reported that the company has had internal talks about acquiring Perplexity to boost its generative AI ambitions.

Apple may be rethinking its AI roadmap, which has lagged behind competitors like Google Assistant and Galaxy AI in recent years. The company already uses OpenAI’s ChatGPT for select Apple Intelligence features.

‍