How Apple developed new ‘Siri AI’ to be more personal and private

Siri understands personal content

During a Q&A session with select media at Apple Park following the keynote, Mike Rockwell, who leads engineering for Siri, demonstrated the new version to journalists and analysts.

Rockwell asked Siri what everyone was bringing to an upcoming potluck lunch, and it gathered the information from separate text messages and displayed it on the screen.

Rockwell then asked Siri which drinks would pair well with the food. Siri accessed Apple’s World Knowledge Service to find relevant suggestions. He then swiped down to expand the interface, which opened the Siri app, where additional information could be viewed by scrolling. The interaction was preserved as an ongoing conversation, allowing users to return to it at any time through the Siri app.

“We did go back and forth on what’s the best way if you want to get back to such a chat that you had,” Apple SVP of Software Engineering Craig Federigh said.

Apple says the new Siri is built on Apple Intelligence and designed to be more personal, conversational, and privacy-focused while remaining deeply integrated across the iPhone, iPad, CarPlay, and other devices. (Image: The Indian Express/ Anuj Bhatia)

“Because you want to continue it, you want to reference it, and quite honestly, the most natural affordance for any user to go find something like that is to have an app that they can manage on their home screen, launch, and get back to. And so we have a Siri app.”

Think of it this way: your iPhone contains your personal information, photos, and much of your digital life. The problem with previous versions of Siri was that it couldn’t use that information effectively to assist you when you needed it.

Now, when you activate Siri AI on your iPhone (also available across Apple devices, including the iPad, CarPlay, and AirPods), simply say, “Hey Siri” or swipe down from the Dynamic Island on your iPhone. Siri can understand your personal context and also perform actions on your behalf across apps, while remaining aware of what’s on your screen through Apple’s Visual Intelligence. Siri will also be able to search the web for answers, much like ChatGPT and Claude.

Siri is also gaining on-screen awareness. It can identify the location shown in a photo you are currently viewing. You can then ask Siri for directions to that location or have it add a stop at a friend’s house along the way. Siri can also proofread text across apps to catch typos and grammatical errors, and it can help you get started by drafting text when you are unsure what to write.

Siri will also serve as a repository for your previous conversations with the assistant, making it easy to revisit and reference them whenever needed.

‘Siri is not a separate chatbot’

While Apple has overhauled Siri AI, making it smarter and more contextually aware, the company is not yet calling the digital assistant a chatbot, as Google does with Gemini and OpenAI’s ChatGPT.

“We see Siri not as a separate chatbot, an unintegrated place you go and chitchat, but rather as an integral but conversational tool that you use in the moment,” said Federighi. “It’s deeply integrated into your experience, understanding what’s on screen… And so while the experiences are conversational, they are really an extension of your system experience, deeply integrated into your flow.”

Apple has partnered with Google to use Gemini models to power the next-generation, more personalised Siri. And Apple isn’t hiding that anymore. In fact, Apple executives at the Q&A maintained that Siri AI is personal and no information is being shared.

Apple executives demonstrate the next-generation Siri at Apple Park in Cupertino, showcasing how the AI-powered assistant can understand personal context, perform actions across apps, and maintain conversational history. (Image: The Indian Express/ Anuj Bhatia)

In a deep dive into Apple’s AI architecture, executives indicated that while the company relies on Google’s AI models, they are not the same Google Gemini available to the public, as many people in the tech industry first thought when the two companies announced a partnership in January.

“We use none of the models that Google deploys to its customers,” Federighi said during a media briefing. “Your requests are completely private to you. They’re never stored. They’re never accessible to anyone.”

“When it comes to the knowledge base, we of course don’t use Google Search or anything like that as the foundation of our system,” he added.

“The Siri app is a great way to get back to a conversation. You can look at what you’ve previously been doing and either continue that conversation or start a new one. But this app isn’t just reaching out to a model in the cloud—it’s built on powerful system software in Apple Intelligence,” Federighi said, adding that when customers make queries about current events, the information will pass through Apple’s in-house system.

‘Maintaining Apple’s unmatched privacy guarantees’

Federighi and his team discussed at length how Apple built Siri AI at the core of Apple Intelligence, acting as a system layer, and how it differs from the AI that others have integrated into their products. Federighi didn’t shy away from making it clear that Apple’s architecture is completely different from Google’s Gemini.

Apple executives then outlined an architecture for the software called a system orchestrator, which routes AI queries to the appropriate model, either on the device or in the cloud, depending on the amount of computing power and personal data required. The system orchestrator is “key to the privacy architecture of our entire system,” said Federighi.

Apple’s AI architecture includes a system orchestrator and Apple’s AFM Core Advanced model. It also includes cloud-based models such as AFM Cloud, AFM Cloud Pro, and ADM Cloud (for images), as well as Apple’s World Knowledge service.

The most powerful of these AI models, known as AFM Cloud Pro, offers quality that’s “similar” to Google’s Gemini frontier models, Apple AI executive Amar Subramanya said. It will run in the cloud on Nvidia GPUs, which are part of Apple’s Private Cloud Compute infrastructure.

During a media briefing at WWDC, Apple executives outlined the architecture behind the new Siri, including on-device intelligence, Private Cloud Compute, and privacy-preserving cloud AI models. (Image: The Indian Express/ Anuj Bhatia)

Apple executives explained that the technology was developed under the company’s Private Cloud Compute philosophy, which now extends to third-party cloud partners, including Google. That approach is designed to protect privacy even when data is in the cloud.

“We work with both Google and Nvidia to extend our private cloud compute infrastructure to Nvidia GPUs in Google’s cloud, while maintaining Apple’s unmatched privacy guarantees,” Subramanya said.

Sebastian Marineau-Mes, Vice President, Software, said that Apple wanted to use Nvidia’s latest chips but wanted them configured in a more private way, so they couldn’t read what was on the servers.

Marineau-Mes said that recent Nvidia improvements, such as a technology called “ambiguous confidential compute,” allowed Apple and Google to build a system that met its standards.

“We wanted to avail ourselves of the latest technology from Nvidia, and so we set out to extend private cloud compute to third-party cloud,” Marineau-Mes said.

Rockwell acknowledged that past attempts to revamp Siri had not met Apple’s standards. “We have always had this vision for Siri to be the most capable and personal assistant possible,” he said.

“Last year, we actually built a first version of this that was sort of incremental on top of the original Siri, adding tool-calling capabilities. We had it working, but we didn’t feel it was really delivering on the vision and experience we wanted. We also had a design that required much more extensive changes, and we decided to go with that.”

An attempt to make AI more ‘personal’ and ‘human’

Apple previously failed to make Siri AI work, and despite its bold claims, the company is back at the drawing board to make its digital assistant perform effectively on the iPhone and other Apple devices.

During the keynote and media briefing, Apple executives said Siri is built on a modern architecture, is natively multimodal, and supports privacy-first interactions across all its products, creating a seamless experience. It can access and act on personal content locally, such as messages, notes, and photos, without compromising privacy.

Even though Apple has introduced a separate Siri app that allows users to continue and reference conversations, the company insists that the assistant is an extension of system workflows rather than a standalone chatbot.

The demos Apple gave of Siri AI looked promising, and compared to other tech companies and their chatbots, Apple’s AI strategy is more rooted in making Siri practical and focused on making common, day-to-day tasks on your devices easier to accomplish. This sets Siri AI apart from OpenAI’s ChatGPT and Google’s Gemini, which at times can feel more robotic and less human.

(The reporter was in Cupertino at the invitation of Apple)