Apple’s big Siri upgrade still doesn’t have a launch date. Why AI can be hard to predict

Apple kicked off its Worldwide Developers Conference with a frank admission on Monday: The advanced version of Siri it announced at the very same conference one year ago still isn’t ready.

“As we’ve shared, we’re continuing our work to deliver the features that make Siri even more personal,” Apple’s senior vice president of software engineering, Craig Federighi, said during the event, echoing comments from CEO Tim Cook on the company’s most recent earnings call. “This work needed more time to reach our high-quality bar, and we look forward to sharing more about it in the coming year.”

In a little more than two years, AI has gone from powering what was once a niche chatbot to being a catalyst for what some tech leaders are calling a tidal wave that could be as life changing as the internet. But Apple’s Siri delay is perhaps another sign that the technology’s evolution is far more difficult to anticipate than previous major technological advancements, like the smartphone and social media.

While Apple and Samsung release new mobile devices once or twice a year, AI models are constantly evolving — and are far less predictable than cyclical products.

“(AI models) are launching far, far faster than once a year. These updates are actually fast and furious,” Oren Etzioni, former CEO of the nonprofit Allen Institute for Artificial Intelligence, told CNN. “…These models can be opaque, unpredictable (and) difficult to measure because they’re so general.”

And the way people embrace AI assistants might look different from how smartphones, web browsers and social media apps have shown up in our lives. In the smartphone and social media industries, major players such as Apple, Google and Meta emerged early on and cemented their position for more than a decade. But in the AI industry, being first may not necessarily always guarantee long-term success.

And that could be good news for Apple.

“It seems like it won’t be a sort of ‘winner-takes-all’ market,” said Daniel Keum, associate professor of management at Columbia Business School.

High-profile AI delays

AI somehow seems like it’s moving lightning fast but also not quickly enough, as evidenced by product delays like Apple’s.

But the iPhone maker isn’t the only one seemingly facing setbacks. OpenAI has yet to release its anticipated GPT-5 model, and Meta is said to have pushed back the launch of its next major Llama model, according to The Wall Street Journal.

Meanwhile, the advancement and adoption of AI are accelerating; performance has significantly improved from 2023 to 2024; and 78% of businesses reported using AI in 2024, up from 55% in 2023, according to Stanford University’s 2025 AI Index Report.

Part of the reason timelines for AI updates seem to be shifting so frequently is that performance can be hard to quantify, says Etzioni. An AI tool might excel in one area and fall behind in another, and a small change may lead to an unpredictable shift in how the product works: In May, ChatGPT became “annoying” after an update and xAI’s Grok chatbot went on unprompted rants about “White genocide” in South Africa.

Tech companies are also still establishing how frequently they’ll be able to release major updates that significantly impact how consumers use AI chatbots and models versus smaller, more incremental updates. That differs from more familiar tech categories like phones and laptops, and even software upgrades like new versions of Google’s Android and Apple’s iOS, which get major platform-wide updates each year.

Changes in AI might see less fanfare because the technology’s advancements “are becoming more incremental and harder to label and to present to customers as really significant changes,” said Leo Gebbie, principal analyst at tech analysis firm CCS Insight.

Not a winner-takes-all market

Tech that defined the early 2000s, like Facebook and the iPhone, heavily benefited from what’s known as “the network effect,” or the idea that the more people use a product, the more valuable it becomes.

Without its massive network of users, Facebook wouldn’t have become the behemoth it is today. Facebook parent Meta says its products are used by 3 billion people worldwide. Apple-exclusive services like iMessage are a major selling point for Apple’s products.

But AI platforms like OpenAI’s ChatGPT and Google’s Gemini are not meant to be social. While these services will likely improve the more people use them, it doesn’t really matter whether a person’s friends or family are using them. AI assistants become more useful as they get to know you.

The fact that people carry their iPhones — along with their AirPods and Apple Watch — everywhere they go could give Apple’s new Siri an advantage whenever it does launch. Apple said last year that the revamped digital helper will draw on personal information based on iPhone activity to provide individualized answers.

“It’s people doing their own individual tasks,” Darrell West, senior fellow at the Brookings Institution’s Center for Technology Innovation, said in reference to AI assistants broadly. “It’s not like the platform becomes more valuable if all your friends are on the same platform.”

AI may also defy the longstanding narrative that being first is best in the technology industry, as was the case in smartphones, social media and web browsers. Apple’s iOS and Google’s Android dominate the mobile device market, marking the end of mobile platforms from the pre-smartphone era like BlackBerry OS, Microsoft’s Windows Phone and Nokia’s Symbian. Google’s Chrome browser accounts for roughly 67% of global browser usage, while even Apple’s Safari is a distant second at about 17%, according to StatCounter GlobalStats.

And Americans tend to remain devoted to their smartphone platform of choice, as iOS and Android both see customer loyalty rates at above 90%, Consumer Intelligence Research Partners reported in 2023.

But it’s unclear whether similar usage patterns will emerge in AI. While users may be inclined to stick with chatbots and services that learn their preferences, it’s also possible people may use multiple specialized services for different tasks.

That takes the pressure off companies like Apple to worry about falling behind, as consumers may not be as tightly locked in to whichever AI service they happen to use first.

“Even if I fall behind, like a quarter generation, I can easily catch up,” said Keum. “And once I improve, people will come back to me.”

This story has been updated with additional context and information.