Apple has been under pressure for months. The accusation is that it has fallen behind in artificial intelligence and promised more with Apple Intelligence than it ultimately delivered. While some of this criticism is justified, it also falls short in other areas. A new report now suggests that Apple is considering a fundamental change in strategy internally. Instead of continuing to rely on its own large language models, the company could abandon development altogether and use external models. Upon closer examination, this approach appears surprisingly sensible.
Apple was deliberately given time initially. This had two clear reasons. Firstly, the company is known for rarely being the first to market. Apple usually waits, observes the mistakes of others, and then tries to deliver a more stable and better solution. Especially with large language models, there were early examples of how problematic hasty releases can be.
Microsoft's early AI chatbot repeatedly made headlines for all the wrong reasons. It insulted users, demanded apologies, suggested problematic content, and exhibited bizarre behaviors in conversations. In one well-known instance, Kevin Roose of The New York Times published a chat in which the bot declared itself to be in love. In another case, the chatbot demanded payment for a translation and even provided a completely fabricated PayPal address. Such incidents clearly demonstrated how risky immature language AI can be.
This is a particularly sensitive issue for Apple, as language is the most important interface for AI within its own ecosystem. Errors here have not only technical consequences, but also immediate personal ones.
Apple's data privacy as a special case
The second reason for patience lay in data privacy. Large language models require enormous amounts of training data. Many companies delve deeply into user data for this purpose. Apple pursues a different approach and has emphasized for years that personal data is not used for training. This commitment is a central part of the brand, but it makes developing its own AI models significantly more complex.
Compared to other providers, this difference is obvious. While competitors made faster progress, Apple lagged behind, not least because of its own strict data protection rules.
Patience has its limits
After roughly two years without any visible breakthroughs, it became clear that simply waiting was no longer a solution. A truly intelligent Siri is now a standard feature of modern smartphones. The idea of allowing users to replace Siri with alternative chatbots would have been a pragmatic interim step.
Such an approach would have had several advantages. The time pressure on Apple would have decreased, as users could have temporarily switched to other solutions. At the same time, Apple would have gained valuable insights into real user queries. This data would have been extremely helpful for the further development of Siri.
Of course, this would have required explicit user consent. The willingness to give it would likely have been high, since a better Siri is in everyone's interest.
But even after another six months without any discernible progress, it became increasingly difficult to believe that Apple could deliver in a timely manner using only its own models.
A possible change in strategy
A recent report suggests that Apple may have internally reached precisely this conclusion. Some executives reportedly believe that large language models will become commonplace in the coming years. Under these circumstances, it seems pointless to invest enormous sums in developing its own models.
There had already been indications that Apple might rely on Google's Gemini for many Siri requests. The crucial detail here is the technical implementation. The specially adapted Gemini model is said to run on Apple's own private cloud compute servers.
Apple has announced that the new Siri will be able to answer personal questions by searching data directly on the device and generating spontaneous answers. Examples include requests such as finding a book recommendation from private messages or notes.
Why this approach makes sense
The idea of using external high-end models is not new. Siri is already configured to access ChatGPT when needed. Apple assures users that these requests are not used for training purposes, unlike direct use of the ChatGPT app.
The Gemini report goes even further. Although the model originates from Google, running it on Apple's own infrastructure ensures complete privacy protection. The data never leaves Apple's controlled environment.
This creates a combination of powerful AI from leading providers and the strict data protection standards for which Apple is known.
Apple's most sensible AI compromise
This approach combines several advantages. Users gain access to modern, powerful AI. At the same time, Apple's data privacy promises are upheld. This is precisely where Apple's real strength in the field of AI lies, not in the sheer power of the models themselves.
Given the slow progress of its own AI models, it makes sense to shift focus. If Apple consistently uses the best available AI models but operates them on its own servers with its own data privacy guarantees, a technically compelling solution emerges that remains true to the brand promise. This could be the decisive step for the future of Siri. The best products for you: Our Amazon storefront offers a wide selection of accessories, including those for HomeKit. (Image: Shutterstock / LittlePigPower)
- Apple is preparing to launch AI-based smart glasses
- MacBook innovations 2026: Three models launch at the beginning of the year
- iOS 27 in detail: New services, AI and focus on stability
- Apple in January 2026: These products and updates are coming
- iPad Air 2026: All the information on technology, design and release
- Apple TV 4K: All the rumors about the big 2026 update
- Apple 2026: An overview of all new devices, leaks, and dates
- Apple 2025: How broad and strong the product year was
- iPhone Fold: These features set new standards in the market
- iPad 12: Details on design, hardware, and the planned launch
- iPad mini 8: Big changes for the small Apple tablet
- Apple launches major Mac offensive in 2026 with new chips
- iPhone 18 Pro: How Apple is further developing the Pro series
- Can Apple's A18 budget MacBook compete in the market?
- Apple TV Hits 2026: These are the releases you shouldn't miss
- Apple's AI strategy: Why the crisis is smaller than expected
- iPhone Fold: Why the outer display is much smaller
- Apple Glasses as a technical intermediate step towards AR glasses
- Apple 2026: Up to 25 new products – here's what to expect
- iOS 26.3: Here's what we know about the release date and content
- Apple iPad Roadmap 2026: These models are planned
- AirPods Pro: Apple plans major changes for 2026
- Buy an iMac or wait: What upcoming models should offer
- Apple Products 2026-2027: Leaks about iPhones, Macs and more
- iPhone Fold: New detail makes a high starting price more likely
