SEOUL, Feb 3, 2026, 18:23 (KST)
- A test on a Samsung Galaxy S24 FE found some “Galaxy AI” tools stopped working locally after AICore was disabled
- Google says AICore supplies Android apps with up-to-date on-device AI models, including Gemini Nano
- Samsung says users can force on-device processing, but many AI functions still rely on the cloud
Android Authority said disabling Google’s AICore system service on Samsung Electronics phones can knock out some on-device AI features and push others back to cloud processing. The site’s editor Andy Walker tested the change on a Galaxy S24 FE running Android 16 and said image tools such as Samsung Gallery’s Generative Edit and Sketch to image only worked again after he switched off “process data only on device”, while Circle to Search and Live Transcribe still ran. He reported no clear speed-up and said the app still took up about 1.3GB of storage after being disabled, though he cited colleagues and online users who reported better battery life on some devices. (Android Authority)
The debate is landing as handset makers keep wiring “AI” tools deeper into the phone, not as optional downloads but as system services. Once they sit there, always available, they become another thing users try to trim when battery life feels tight.
It also flips a simple privacy story into a trade-off. Remove local AI capability and the phone may try to do the same work elsewhere, which can mean sending text or images over the internet instead of keeping them on the device.
On Google Play, Google describes Android AICore as powering features “across Android” and supplying apps with the latest AI models. The listing’s data-safety section says the app does not share data with third parties, but may collect app-performance information and device identifiers, with data encrypted in transit. (Android Authority)
Google’s Android developer documentation says its Gemini Nano model — a smaller version of Gemini designed to run on phones — operates inside Android’s AICore system service, using device hardware to keep inference (the model’s response generation) fast and to keep the model up to date. The same page points developers to APIs for summarisation, proofreading, rewriting, image descriptions and speech recognition without needing a network connection. (Android Developers)
In a 2023 post on the Android Developers Blog, Dave Burke described Android AICore as a system service in Android 14 that gives apps access to Gemini Nano while handling model management and safety features. Burke wrote that AICore is “private by design” and isolated from the network, pointing to Android’s Private Compute Core as a model for how sensitive processing can be separated from online access. (Android Developers Blog)
Samsung says its Galaxy AI features split between on-device and cloud-based work, and that users can disable cloud processing with a “master switch” in Galaxy AI settings labelled “Process data only on device.” The company adds that most translations run on the device, but “most other features require a network connection and will utilize the cloud in some way.” (Samsung)
For companies managing phones at scale, the control points are already documented. Microsoft’s Intune guidance describes AICore as an on-device AI system app and lists the package name “com.google.android.aicore” for administrators who want to disable it on Android Enterprise devices. (Microsoft Learn)
But users chasing battery savings or privacy can end up with a different risk profile. Samsung’s Knox enterprise documentation says Galaxy AI features processed on-device don’t use user data for training, while some cloud-processed features may be used for model training — a reminder that shifting work off the phone can change where data goes, and what it is used for. (Samsung Knox)
The bottom line is uneven. For people who rarely touch AI features, turning off AICore may feel like dead weight removed. For others, it can quietly turn an offline tool into an online one, or leave a feature greyed out until settings are flipped back.
