Google's on-device AI efforts irk consumers over shrinking storage space

Google's growing push toward on-device AI is reportedly consuming gigabytes of storage on Android phones and PCs, raising concerns over transparency and user control

Some Google services are reportedly consuming storage space on smartphones and PCs without users explicitly realising it. Recent reports by 9To5Google and Engadget highlighted how Google’s growing use of on-device artificial intelligence models can temporarily occupy several gigabytes of storage in the background.

 

In some cases, users reportedly found large AI-related files downloaded automatically without explicit prompts. At a time when NAND and DRAM prices are rising, this additional storage usage is effectively consuming paid storage capacity without many users being fully aware of it.  But why is this happening, and what exactly is being stored? 

 

Chrome is reportedly consuming 4GB of storage on PCs

According to an Engadget report, computer scientist Alexander Hanff recently alleged that Chrome was downloading a large AI-related file called “weights.bin” in the background without clearly seeking user consent.

 

The file reportedly exceeded 4GB in size and was linked to Gemini Nano, which powers Chrome’s on-device AI features.

 

Engadget reported that the file appeared inside Chrome directories on some macOS and Windows systems after recent browser updates.

 

The report also noted that deleting the file sometimes did not solve the issue, as Chrome could reportedly download it again automatically.

 

This has raised concerns around transparency, privacy regulations, and even the environmental impact of distributing large AI models to millions of devices globally.

Google’s explanation for Chrome downloads

Google said the storage usage is linked to Gemini Nano, which has been available in Chrome since 2024 as a lightweight on-device AI model used for features such as scam detection and developer APIs.

 

According to Google’s support documentation, Chrome downloads these on-device generative AI models in the background so AI-powered features remain ready for use. The company said the models may support tools such as writing assistance, scam warnings, webpage summarisation, and tab organisation.

 

Google also responded to claims regarding files reappearing after deletion. The company said users can manually disable on-device AI models through Chrome’s settings menu under the “System” section. Turning the feature off removes locally stored AI models, though AI features that depend on them will stop functioning until the models are downloaded again.

 

Google further stated that the models may automatically uninstall themselves if a device runs low on resources, including storage space.

What is consuming extra space on Android smartphones

Android AICore is Google’s system service that allows generative AI features to run directly on Android devices instead of relying entirely on cloud servers. According to Google’s support documentation, the service is available on supported devices running Android 14 and above. AICore manages Gemini Nano, Google’s lightweight AI model designed specifically for smartphones and tablets.

 

The company said this allows AI-powered features to function locally on-device, improving privacy, reducing latency, and enabling some tools to work even without an internet connection.

Google said AICore powers features such as advanced proofreading, smart replies in apps like WhatsApp, scam detection, audio transcription, text summarisation, and translation.

 

According to a report by 9To5Google, the service can sometimes consume more storage than expected without notifying users.

Why AICore sometimes takes more space

According to the report, the increased storage usage is linked to how Android handles Gemini Nano updates. Google later explained the reason in an official support document. The company said that when a new version of Gemini Nano becomes available, Android temporarily keeps both the old and new AI models stored on the device for up to three days.

 

Google described this as a fail-safe mechanism that allows devices to instantly roll back to an older model if the update encounters issues, instead of downloading large files again. However, the increase in storage consumption is temporary. Once the system confirms the update is stable, the older model is automatically removed and the additional storage space is cleared.

 

Google also stressed that on-device AI processing is intended to improve privacy because sensitive data remains stored locally on the device instead of being sent to cloud servers.

Why Google is pushing on-device AI

The broader push toward on-device AI reflects a wider industry trend where companies are increasingly trying to run AI systems locally instead of depending entirely on cloud infrastructure.

Running AI models directly on devices can improve response times, reduce server costs, and offer stronger privacy protections because user data does not always need to leave the device.

 

However, the trade-off is that these models can occupy significant local storage and may run in the background without many users fully realising it.

 

As companies continue embedding AI deeper into operating systems and browsers, users are likely to see more local AI models quietly becoming part of everyday software experiences.