Why We’re Not Getting That AI-Powered Siri Anytime Soon

If you were under the assumption that Siri was soon going to be supercharged with AI, you wouldn’t be alone. In fact, Apple has advertised as much since last WWDC, showing off its ChatGPT-like assistant in commercials and promotional materials.

It’s been nearly a year since WWDC 2024, and that new Siri is still not here. The thing is, it likely won’t be for a long time. How long is anyone’s guess (I’ve been tracking the delays here), but one thing seems likely: Apple is probably not showing off AI Siri at next week’s WWDC 2025.

Apple’s AI program is a mess

In a report from Bloomberg’s Mark Gurman last month, he describes a chaotic situation regarding Apple’s AI department. The piece is a fascinating and in-depth look at Apple’s AI woes, and I won’t give a detailed summary of the entire article. However, I will briefly discuss what’s going on, and how it related to AI Siri.

Executive leadership at Apple, including senior vice president of Software Engineering Craig Federighi, didn’t believe AI was worth the investment, and didn’t want to allocate resources away from Apple’s core software components in order to develop the technology. But once Federighi used ChatGPT following its late 2022 launch, he did a 180. He and other Apple executives began meeting with the big AI companies to learn everything they could, and pushed for iOS 18 to have “as many AI-powered features as possible.”

While Apple’s AI department already existed before this scramble (the company had poached Google’s artificial intelligence chief John Giannandrea), the engineers simply couldn’t match the quality or accuracy of the tools provided by other companies like OpenAI, Anthropic, or Google, who had a huge head start on generative AI.

That lag manifested in two ways: First, many of the AI features that Apple did bring to market were half-baked. Apple Intelligence’s notification summaries feature, for example, infamously made some major mistakes, such as “summarizing” a BBC news alert to say that United Healthcare shooting suspect Luigi Mangione had shot himself. (Apple later disabled the feature for news alerts.)

Second, because Apple couldn’t rely on its own tech to carry Apple Intelligence, they outsourced some tech to another AI company. While there was much debate about which company they should work with (Giannandrea wanted to work with Google and bring Gemini to iOS), Apple eventually settled for ChatGPT—which is why OpenAI’s bot is built into your iPhone today.

Apple’s lack of AI focus meant they missed the rush to acquire GPUs—the main processing unit used for training and running AI models. They also have strict privacy policies when it comes to user data, which severely limits what data they can use to train their models. (Some might say that’s actually a good thing, and pause to think about the companies that do have a plethora of data to train with.)

While Apple was able to get some AI features working “well” enough to ship, Siri was never among them. In order to bring AI to Siri, the company had to split Siri’s “brain” in two—one featuring the existing code, used for traditional Siri tasks like setting timers and making calls, and the other for AI. While the AI side in a vacuum can work, integrating it with the other half of Siri’s brain is problematic, and is the cause of much of the delay.

But rather than wait until Apple figured out how to get AI Siri working to actually show off its new features, the company went ahead and heavily marketed them. During WWDC 2024, we saw prerecorded demos of Siri taking complex requests and generating helpful answers by accessing both a knowledge base about the user in question, as well as an awareness of what was happening on-screen. A prime example was an Apple employee asking Siri about their mom’s travel itinerary: Siri dove through the employee’s messages with their mom to draw up the plans.