
A mobile app is several layers, not one technology, and in 2026 the interesting change is where AI runs. The layers are the app itself (native or cross-platform), a back end and data store, security, analytics and the store release process. For AI features you now have a real choice: Apple and Google both let apps use a model on the device, which works offline and keeps data on the phone, while larger cloud models handle harder tasks. Most apps combine the two. Pick each layer for your team and your users, not for fashion.
Key takeaways
- Apple's Foundation Models framework, announced on 9 June 2025, gives apps access to the on-device Apple Intelligence model, free of cost, available offline and privacy-protecting, "with as few as three lines of code" in Swift.
- On Android, Gemini Nano runs in the AICore system service and is reached through ML Kit GenAI APIs; Google lists token limits and device hardware limits, and suggests a hybrid fallback to cloud models.
- GitHub's Octoverse 2025 found Python the second most used language and the leader among AI-tagged repositories, with 582,196 repositories, up 50.7% year over year.
- OWASP's MASVS groups mobile app security into eight areas, from storage to privacy.
- Store rules apply to every approach: Xcode 26 and an iOS 26 SDK for Apple uploads since 28 April 2026, and Android 16 (API level 36) targeting for new apps and updates on Google Play from 31 August 2026.
What layers make up a mobile app?
| Layer | What it does | Typical choices |
|---|---|---|
| The app | The interface and device features | Swift and Kotlin natively, or Flutter, React Native or Kotlin Multiplatform |
| Back end and APIs | Accounts, business rules, integrations | Node.js, Python, Java and others |
| Data | Stores and serves information | A relational or document database, plus file storage |
| AI | Understands text, images or speech, or generates content | On-device models, cloud models, or both |
| Security and privacy | Protects data and accounts | A standard such as OWASP MASVS, plus reviews |
| Release and measurement | Gets the app to users and shows what they do | Store pipelines, crash reporting and analytics |
Which language or framework for the app?
Decide this first, because it shapes the rest. Native Swift and Kotlin give the closest fit to each platform; Flutter, React Native and Kotlin Multiplatform share most of the code between iOS and Android. We compare them, with current versions and store requirements, in Cross-platform or native in 2026.
What about the back end?
Choose the back end for your team and the work it must do. Node.js with TypeScript suits teams that already write JavaScript; see Node.js frameworks in 2026. Python is a common choice where AI and data work are central: GitHub's Octoverse 2025 ranks it second by contributors, behind TypeScript, and found it leads AI-tagged repositories with 582,196 repositories, up 50.7% year over year. Java remains common in large organisations. Whatever you choose, keep it behind an API the app calls, so you can change it later. Our Node.js development and Python development pages describe how we work.
Where should AI run: on the device or in the cloud?
Both platforms now let apps use a model on the phone.
- Apple. The Foundation Models framework, announced on 9 June 2025, gives developers "access to the on-device Apple Intelligence foundation model". Apple says it protects privacy, works offline and is "free of cost", with native Swift support and access "with as few as three lines of code".
- Android. Gemini Nano "lets you deliver rich generative AI experiences without needing a network connection or sending data to the cloud". It runs in Android's AICore system service and is reached through ML Kit GenAI APIs for tasks such as prompting, summarisation, proofreading and rewriting. Google notes on-device constraints such as token limits, device hardware limits and model capability boundaries, and recommends a hybrid fallback to cloud models when Gemini Nano is unavailable on a device.
| On-device model | Cloud model | |
|---|---|---|
| Privacy | Data can stay on the phone | Data is sent to a server |
| Offline use | Works without a network | Needs a connection |
| Cost per use | Low or none to the app | Usually charged per request |
| Capability | Limited by device and model size | Larger, more capable models |
| Device coverage | Only supported devices | Any device with a connection |
A sensible pattern is to use the on-device model for simple, private tasks such as short summaries or rewriting, and send harder tasks to a cloud model with a clear privacy notice and a fallback when there is no network. For the design side of AI in an app, see UX and UI design trends in 2026, and for privacy see Privacy for AI features: Canada and India.
What security and privacy basics apply?
OWASP's Mobile Application Security Verification Standard (MASVS) groups mobile security into eight areas: storage, cryptography, authentication and authorisation, network communication, platform interaction, code quality, resilience against reverse engineering and tampering, and privacy. Use them as a checklist when you brief a team and when you review a release; our hiring guide, How to hire an app development company in 2026, shows how.
What do the stores require?
Apple's page says: "Apps uploaded to App Store Connect must be built with Xcode 26 or later using an SDK for iOS 26, iPadOS 26, tvOS 26, visionOS 26, or watchOS 26", which has applied since 28 April 2026. Google Play requires that, from 31 August 2026, new apps and app updates target Android 16 (API level 36) or higher, with exceptions for some device types, and extensions to 1 November 2026 are available. Budget for these changes every year.
A checklist for choosing your technologies
- Name the day-one features, and which need device hardware.
- Pick native or cross-platform from your team's skills and your design needs.
- Choose the back end for the work it must do, behind a clean API.
- Decide which AI tasks can run on the device and which need the cloud, with a fallback.
- Review against the eight MASVS areas.
- Plan yearly store-required upgrades and maintenance.
- Measure what users do from the first release.
For help, our app development page covers what we build, and our iOS and Android pages cover native work.
FAQ
Which programming language is best for mobile apps?
There is no single best. Swift for iOS and Kotlin for Android give the closest platform fit; Dart (Flutter), JavaScript or TypeScript (React Native) and Kotlin (Multiplatform) share code across both. Choose by your team's skills and the app's needs.
Can an app use AI without an internet connection?
Yes, on supported devices. Apple's Foundation Models framework and Android's Gemini Nano both run models on the device, so some features can work offline. Their capability is limited compared with large cloud models.
Should the AI run on the phone or on a server?
Often both. Use the device for simple, private tasks and a server for harder ones, with a fallback if the device or network does not support the feature.
Do we need a back end for every mobile app?
Not always, but most apps with accounts, payments or shared data do. Keep it behind an API so the app and back end can change independently.
How often do we need to update the app?
At least yearly, to meet Apple's and Google's new build and target requirements, and more often to fix bugs and security issues. Plan a maintenance budget.
Sources (6 links)
- Apple Newsroom, "Apple supercharges its tools and technologies for developers" (9 June 2025): https://www.apple.com/newsroom/2025/06/apple-supercharges-its-tools-and-technologies-for-developers/
- Android Developers, "Gemini Nano": https://developer.android.com/ai/gemini-nano
- GitHub Blog, "Octoverse: A new developer joins GitHub every second as AI leads TypeScript to #1" (28 October 2025): https://github.blog/news-insights/octoverse/octoverse-a-new-developer-joins-github-every-second-as-ai-leads-typescript-to-1/
- OWASP, "Mobile Application Security Verification Standard (MASVS)": https://mas.owasp.org/MASVS/
- Apple Developer, "Upcoming Requirements": https://developer.apple.com/news/upcoming-requirements/
- Android Developers, "Meet Google Play's target API level requirement": https://developer.android.com/google/play/requirements/target-sdk



