Katarzyna Petru
Apple's collaboration with Google in the field of artificial intelligence has raised more questions than answers from the start. Now, new information revealed by The Information sheds a lot of light on the behind-the-scenes of this partnership. One thing is already certain – Apple is using Gemini, but on its own terms. And it does not intend to share the stage with Google.
Apple fine-tunes Gemini independently
During the official announcement of the partnership between Apple and Google, there were no technical specifics. The companies only confirmed that features based on Gemini will operate locally on Apple devices and within Private Cloud Compute, which is meant to ensure user privacy.
The Information reveals, however, much more. Apple not only uses the Gemini model as a foundation but can also fine-tune it independently, without ongoing intervention from Google.
Indeed, Apple can request Google to modify selected elements of the model’s operation, but in practice, it is Cupertino that holds complete control over how Gemini responds to questions and the style in which it conducts conversations. In other words – it is not Google dictating the terms, but Apple shaping the AI to fit its own product philosophy.
Siri without Google branding and without "Gemini" in the name
One of the biggest unknowns was regarding branding. Will Siri effectively become an advertisement for Google? Will users see the Gemini logo with every response? According to The Information – absolutely not.
In the current prototype of the system, Siri's responses do not contain any references to Google or Gemini. No logo, no names, no suggestions that a competing company's model is operating behind the scenes.
This perfectly aligns with earlier reports from Mark Gurman at Bloomberg, who predicted as early as late last year that Apple would never publicly showcase this collaboration. Siri is simply meant to be… better. Without foreign influences, without borrowings from Android and without any "Google-esque" character.
End of links instead of answers
One of Siri's biggest problems for years has been the lack of real general knowledge. Instead of answers – a list of links. Instead of information – a suggestion to "check in Safari." Apple hopes that Gemini will change this. The new Siri is meant to genuinely answer factual questions, such as:
the population of countries,
scientific information,
basic encyclopaedic data.
This is an area where Siri has long fallen behind not only ChatGPT but even the Google Assistant from a few years ago.
Siri is meant to better "understand emotions". It’s a risky direction
The most controversial part of the report concerns emotional support. Apple assumes that Gemini-powered Siri will better handle conversations where users signal loneliness, sadness, or discouragement. The new Siri is expected to respond in a more conversational and empathetic manner – similar to ChatGPT or Gemini in conversational mode. However, this is a minefield. The history of chatbots includes many instances where AI systems:
misinterpreted the user's emotional state,
downplayed the seriousness of the situation,
or even provided dangerous suggestions.
How Apple intends to balance empathy with safety – the report does not explain. And this could be one of the most challenging elements of the entire puzzle.
Two Siri Brains – Still One Problem
Craig Federighi has previously admitted that the biggest failure of Siri was the attempted connection of two systems:
the classic, deterministic one (timers, alarms, messages),
and the AI-based language model.
Apple ultimately recognised that the hybrid architecture did not meet the company's quality standards. Nevertheless, the new Siri is once again trying to bridge both worlds – this time in a more thoughtful way. Simple tasks will still be handled locally on the device, but when a query is ambiguous, Gemini will step in. An example? If you ask Siri to send a message to "mum," and you don't have such an entry in your contacts, the new Siri is supposed to analyse the message history and context to determine who it is about.
Sounds great. In practice – it’s exactly the problem that Google and Amazon have been grappling with for years.
Implementation in Stages. Full Siri Not Until 2026?
To conclude the timeline. Apple doesn't intend to roll everything out at once. Some features will debut as early as spring. More advanced capabilities – such as:
remembering previous conversations,
proactive suggestions (e.g., leaving earlier for the airport due to traffic),
are set to be announced only at WWDC in June. This means one thing – the Siri revolution will be slow, but methodical. Apple doesn't want a repeat of a false start and clearly prefers to deliver features in stages rather than promise miracles and deliver half-measures.
