Where the Rumor Originates

In March 2024, Bloomberg reported that Apple engaged in discussions with Google to license Gemini, which would help power some AI features on the iPhone, including a part of the much-anticipated “Siri 2.0” upgrade. These talks reportedly centered on utilizing a cloud-based large model for more complex tasks, while Apple would continue to develop its own on-device models for tasks that require privacy. Apple has also explored collaborations with OpenAI and others.[Bloomberg]

Since then, Apple’s public announcements provide some useful context. During WWDC 2024, Apple introduced Apple Intelligence, a hybrid AI system that uses smaller on-device models for many tasks while tapping into larger models in the cloud as needed, all with a focus on privacy through a feature called Private Cloud Compute. Apple also revealed optional access to ChatGPT on iOS, iPadOS, and macOS for specific queries.[Apple][Apple Security]

Taken together, it’s clear that Apple is aiming for impressive AI features on the iPhone. It plans to use its own models wherever possible, selectively bringing in third-party models for more challenging problems. Gemini could become one of those third-party partners for Siri 2.0 if an agreement is reached.

What Gemini Brings to the Table

Gemini is Google’s family of multimodal AI models designed to understand and process text, images, audio, and code. Google touts strengths such as long-context reasoning (notably in Gemini 1.5 Pro), multimodal understanding, and tool utilization, all of which are pertinent for a modern assistant.[Google]

• Multimodal Input and Output: Interpret text, images, and more all in the same request.
• Longer Context Windows: Maintain more of a conversation for follow-up questions.
• Tool Use and Function Calling: Trigger actions like retrieving files or opening apps through APIs.
• Developer Ecosystem: Extensive support across Google Cloud and Android may lead to rapid iterations.

On Android, Google has begun integrating Gemini into consumer experiences, partnering with phone manufacturers like Samsung on features such as Circle to Search.[Google]

How Siri 2.0 Could Use a Partner Model Like Gemini

Apple has indicated a layered approach to AI on the iPhone:

• On-Device Models: Handle many personal and context-rich tasks locally for speed and privacy (like summarizing notifications or rewriting text).
• Private Cloud Compute: When a larger model is required, requests may be sent to Apple-controlled servers, designed with end-to-end encryption and limited data access.[Apple Security]
• Partner Models: For specific prompts, iOS could offer to route the request to an external provider like ChatGPT (or potentially Gemini) with clear consent and data-use disclosures.[Apple]

If a deal between Apple and Google materializes, here’s how Gemini might be integrated within Siri 2.0:

• Routing Complex Queries: When on-device models lack confidence, Siri could seek permission to send the request to Gemini for more accurate answers.
• Specialized Skills: Gemini might assist with long-context research summaries, multimodal analyses, or advanced planning that exceeds what on-device models can handle.
• Developer Actions: With user consent, a partner model could trigger app shortcuts or suggest automations through Siri intents.
• Fallback and Choice: Users may be able to select a preferred partner model, similar to the existing option for ChatGPT.

It’s important to note that as of late 2024, Apple has only confirmed optional ChatGPT integrations. Any Gemini integration remains speculative and should be approached with caution.[The Verge]

Privacy, Security, and User Control

Privacy is at the center of Apple’s AI offering. With Private Cloud Compute, Apple asserts that requests for cloud-scale models are processed on servers it controls, utilizing verifiable software images and limited data access. This design is aimed at safeguarding personal data, even when tasks extend beyond on-device capabilities.[Apple Security]

Introducing a third-party model like Gemini raises additional considerations:

• Consent and Transparency: Users should expect clear prompts before any request is sent to an external provider, including links to the provider’s data policies.
• Scoping and Minimization: Only the necessary parts of your request should be shared, and only for as long as is required to complete it.
• Account Choice: Users might choose to interact with Gemini anonymously or log in for personalized features, echoing Apple’s approach with ChatGPT.
• Granular Controls: Options to keep certain data strictly on-device could be offered per app or feature.

In summary, it’s likely that Apple will maintain conservative privacy defaults and require explicit opt-ins for any third-party interactions.

How This Fits Into the Competitive Landscape

The landscape for AI assistants is evolving quickly:

• Google Gemini is already powering consumer and developer experiences across Android and the web, showcasing its long-context and multimodal capabilities.[Google]
• Samsung is leveraging Google’s AI and its own features to enhance Galaxy AI capabilities on its latest devices.[Samsung]
• Microsoft integrates Copilot across Windows and Office, focusing on productivity and enterprise solutions.[Microsoft]
• Apple’s Apple Intelligence combines on-device models with Private Cloud Compute and optional ChatGPT integration, with planned incremental Siri upgrades.[Apple]

In this context, adding Gemini as a partner would not represent a significant philosophical shift for Apple, but rather a practical approach to delivering top-notch results across a broader range of tasks.

What This Could Mean for You

• Better Answers, Fewer Dead Ends: Complex or open-ended inquiries could be routed to a more capable model when necessary.
• Smarter Automations: Imagine multi-step tasks like, “Plan a 3-day trip to Seattle, then create a packing list and add it to Reminders,” with information drawn from your calendar, subject to your permissions.
• More Reliable Follow-ups: Long-context models could maintain conversational threads, so you won’t need to repeat yourself.
• Clearer Controls: Look forward to straightforward options to maintain data on-device, approve model routing, or choose a preferred provider.

Open Questions to Watch

• Scope: Will Gemini assist only with certain Siri queries, or also integrate into features like Photos, Notes, and Safari?
• Regions and Languages: Apple typically implements AI features gradually by language and region. A partner model might impact this rollout schedule.
• Costs: Will any functionalities require a subscription or sign-in with a third-party account?
• Developer Access: Will app developers have the ability to request partner-model routing through Siri intents or App Intents?
• Timing: Apple usually previews major Siri changes at WWDC, rolling them out gradually in subsequent iOS updates. The exact timing for any Gemini-related features remains unknown until officially announced.

The Bottom Line

Apple adopting a multi-model strategy is not unexpected. The goal is to handle as much as possible on-device, leverage Apple-controlled cloud services when necessary, and then provide optional access to external models for complex prompts. If a deal with Gemini goes through, Siri 2.0 could become more capable and adaptable without compromising Apple’s commitment to privacy. Until Apple makes an official announcement, treat these reports as credible but unconfirmed and keep an eye out for signs in future iOS updates and developer documentation.

FAQs

Is Apple Definitely Adding Google Gemini to Siri?

No. As of late 2024, Bloomberg reported talks between the two companies, but Apple has not confirmed Gemini integration. Apple has announced optional ChatGPT access and its own Apple Intelligence platform.

How Would Gemini Help Siri 2.0?

Gemini could manage complex, long-context, or multimodal queries that exceed the capabilities of on-device models, providing smarter answers or better planning. Routine and sensitive tasks would continue to be handled locally or via Apple’s Private Cloud Compute.

Will My Data Be Shared with Google if This Happens?

Apple states that third-party model access is opt-in and clearly communicated. If routing to a partner model occurs, you should receive prompts and controls explaining what is shared and how.

What About ChatGPT on iPhone?

Apple offers optional access to ChatGPT within iOS, iPadOS, and macOS for certain queries. A potential Gemini option would likely be available alongside, rather than as a replacement for, that functionality.

When Will Siri 2.0 Arrive?

Apple typically reveals major Siri upgrades at WWDC and rolls them out in stages through iOS updates. The specific timing for any Gemini-related features is still unknown until officially announced.

Sources

1. Bloomberg – Apple and Google in Talks to Let Gemini Power iPhone AI Features (Mar 2024)
2. The Verge – Apple Reportedly in Talks to Bring Gemini AI to iPhone (Mar 2024)
3. Apple Newsroom – Introducing Apple Intelligence (Jun 2024)
4. Apple Security Research – Private Cloud Compute (Jun 2024)
5. Google – Gemini 1.5 Update and Long-Context Capabilities (Feb 2024)
6. Google – AI on Android with Samsung Galaxy S24 and Circle to Search (Jan 2024)
7. Microsoft – What’s New in Copilot (Sep 2024)