Apple’s surprise news this week signals a huge shift for iPhone—it almost doesn’t matter what happens next, it looks like everything has already changed…
It’s the smartphone news of the week—Apple is reportedly talking to Google about a blockbuster deal to adopt Gemini for iPhone. According to Bloomberg, which broke the story, this “blockbuster agreement would shake up the AI industry.”
This news is—of course—far from certain and unconfirmed, and currently relies on “people who asked not to be identified because the deliberations are private.” But it’s an update that has set the tech world alight at just what this could mean.
In reality, it doesn’t matter whether this consummates or not—something on Apple’s side has changed and this is the real update for iPhone users. Outside its ecosystem, the generative AI world is dominated by Google with Gemini (née Bard) and OpenAI with ChatGPT, co-opted by Microsoft. It’s therefore unsurprising that “Apple also recently held discussions with OpenAI and has considered using its model.”
Apple has teased that its forthcoming iOS 18, due as ever in the fall, will be the “biggest” update ever—cue loud music and applause. And that’s all about on-device AI coming to iPhone. According to reports some weeks ago, Apple has been testing its own on-device large language models against OpenAI—and others—to ensure it’s commensurate with cloud offerings, while maintaining its edge focus.
But the news that there might be a deal with Google—or OpenAI—suggests that either Apple is not as well progressed as thought, or—perhaps more likely—that there are intrinsic limitations for on-device generative AI versus what can be done in the cloud.
The suggestion is that Apple may split its AI offering, with privacy-centric capabilities device-side—maybe using Gemini Nano, and heavier lift actions such as writing or analyzing long form documents, or conducting detailed research, in the the cloud.
Over to Samsung. The Android handset giant has focused its marketing machine on Galaxy AI this year, as it takes on both Google and iPhone in the high-profile and soon to be high-growth AI handset market. It had seemed that Apple’s in-house, on-device focus would differentiate to Samsung’s hybrid device/cloud model, both for its own offerings as well as those it takes from Google under the Android umbrella.
“When integrated into phones, AI is, to put it simply, a revolution,” Samsung’s MX lead said ahead of MWC. “It’s also important to raise the standards of security and privacy in this new era of data-intensive mobile experiences. That is one of the reasons we’ve taken a hybrid approach that combines on-device and cloud-based AI. Besides ensuring seamless usability, this lets users limit some features to function entirely on-device, giving them greater control over what they do with their data.”
At the time, this seemed to signal a difference to where Apple would head. But now, if the new reports are correct, then not so much. The very fact Apple is exploring the same hybrid approach is a significant indicator as to where this all goes next.
The issue for Samsung’s AI offerings has always been that mix and match between its own capabilities and Google’s. But again, these reports suggest this might replicate the approach Apple takes, narrowing the iPhone/Android gap even further.
With little performance differences between top-end flagships from Apple and Samsung—which account for the top selling premium handsets globally, and the sales driver AI is seen as becoming, this could be the closest the two ecosystems have ever been, as Samsung becomes more like Apple, and Apple does the same.
But Apple—more than Samsung—has a key challenge. AI is the next privacy battlefield. We are already seeing reports that generative AI prompts can be hacked and it’s seen as when not if we see major breaches.
Some companies regularly scan personal information in the cloud,” Apple said last month, “to monetize the information of their users. Apple does not. We have chosen a very different path—one that prioritizes the security and privacy of our users.” That was seen as a likely barb at Google and its cloud offerings versus Apple’s own.
Tools of mass surveillance,” Apple further warned in its response to proposed surveillance legislation in Australia, “have widespread negative implications for freedom of opinion and expression and, by extension, democracy as a whole. For example, awareness that the government may compel a provider to watch what people are doing raises the serious risk of chilling legitimate associational, expressive, political freedoms, and economic activity.”
This has implications for cloud versus on-device AI, especially given that anything sent to AI in the cloud cannot—by its very nature—be end-to-end encrypted. That means it’s open to storage and retrieval and to legislation and law enforcement. We have seen this issue with Google Messages becoming a UI for Gemini, but with a warning that it’s outside the platform’s usual full encryption.
The dividing lines between on-device and cloud AI, and how Apple’s approach differs to its rivals, is about to become critical. In some ways this update just mirrors current search arrangements, where Google is already the default across iPhone and Android. But we all know AI is different and those lines will blur. As Apple potentially narrows the gap to Samsung and Google, it will need to tread very carefully.