Apple says its revamped Siri, called Siri AI, will launch in beta later this year after two years of delays, marking a major step in the company’s push to make its voice assistant more useful in the era of artificial intelligence. Unveiled at Apple’s Worldwide Developers Conference in Cupertino, California, the new assistant is designed to go beyond basic commands and act more like a digital helper that can understand context, retrieve information across apps, and surface relevant content from a user’s device or the web.
The update is intended to move Siri away from its long-standing role as a simple task manager and toward a more capable, personalized assistant. Apple demonstrated early examples showing Siri AI finding information from messages and apps without requiring users to switch between services. In one example, the assistant could identify the name of a podcast mentioned by a family member in a text message and present it directly to the user.
Apple said the new Siri experience is built on Apple Intelligence and reflects the company’s broader vision for an AI-powered assistant that understands what users want and when they want it. The company first previewed this direction at WWDC 2024, and the latest demos suggest it is now closer to reality. However, Apple’s progress also appears to have depended in part on outside help, especially from Google.
According to the report, Apple and Google reached an agreement in January allowing Apple to use Google’s Gemini models to power Siri’s upgraded capabilities. Apple software chief Craig Federighi said during a WWDC question-and-answer session that Siri AI does not use Google’s Gemini assistant directly, nor does it rely on Google Search or the same infrastructure Google uses for its own customers. Instead, Apple uses Google’s models as part of a different system built for its own products.
Apple vice president of AI Amar Subramanya said the company built its Apple Frontier Models using proprietary data and reinforcement learning, then refined them with outputs from Google’s Gemini Frontier models. That helped Apple develop a stronger AI foundation after its in-house efforts reportedly did not reach the performance needed for the new Siri.
The company is also expanding its private AI infrastructure. Apple said Siri AI will run across on-device models and Private Cloud Compute, its secure server system designed to process requests without exposing user data to Apple or storing it for future use. But Apple’s cloud setup did not have enough power for its most advanced model, so the company worked with Google and Nvidia to extend its Private Cloud Compute system to Nvidia chips hosted in Google Cloud.
Apple says it retains strict control over the software running on those Nvidia-powered Google servers, with only Apple-signed software allowed to communicate with Apple devices. The company says this is meant to preserve privacy while giving Siri the power needed for more advanced AI tasks.
Siri AI is expected to arrive this fall.
