At the Worldwide Developers Conference 2024, Apple presented the new AI platform “Apple Intelligence” – and the showpiece was to be a new generation of Siri with AI support. But nothing came of it: at the beginning of 2025, Apple had to announce that the introduction would be postponed to 2026. During this time, the internal developments probably did not go as planned, because at the beginning of 2026 the company finally announced that it wanted to use Google Gemini instead of its own development. But the effort to integrate Google Gemini now also appears to be plagued by difficulties, as Bloomberg’s Mark Gurman reports.
Delivery across multiple versions
Apple originally aimed to finally be able to deliver the new functions to customers with iOS 26.4 and macOS 26.4. The new versions should appear in the final version in March or April 2026. But according to Gurman, some developments are behind schedule – and Apple now wants to gradually deliver the features across several operating system versions. Some new functions will be activated with iOS 26.5, which will probably be released in the summer – but others will not be activated until iOS 27 in autumn 2026.
Access to personal data is the biggest problem
In the summer of 2024, Apple demonstrated at WWDC how Siri evaluates the user’s personal data and analyzes messages, emails and calendar entries to assist the user with daily tasks. But according to Gurman, it is precisely this very useful function that is causing major problems – and could come onto the market much later. Furthermore, the development of possible voice control of apps is also said to be problematic.
Public failure
In the company’s recent history, it has rarely happened that previously announced projects were significantly delayed or even failed. Artificial intelligence is currently the main topic in the IT world – and market observers and customers agree that Apple has lost touch with the competition. The Apple world received very mixed feelings about the fact that the company had not managed to develop competitive AI models internally and had to resort to Google Gemini instead.
