Google Chrome silently downloads 4GB AI model without user consent, sparking privacy concerns

Reviewed byNidhi Govil

11 Sources

Share

Google Chrome has been automatically downloading a 4GB Gemini Nano AI model to user devices without clear notification or consent, researchers report. The weights.bin file powers on-device AI features but raises questions about storage consumption, bandwidth costs, and potential violations of European privacy laws including GDPR.

Google Chrome Downloads 4GB AI Model Without Clear User Notification

Source: Lifehacker

Source: Lifehacker

Google Chrome has been automatically installing a 4GB AI model on user devices without explicit consent, according to reports from security researcher Alexander Hanff and multiple user confirmations. The file, named weights.bin, contains training parameters for Gemini Nano, Google's lightweight on-device language model that powers various Chrome AI features

1

2

. Users discovering unexplained drops in available storage space are now finding this substantial file hidden deep within their Chrome directory, stored in a folder called OptGuideOnDeviceModel

3

.

Source: PCWorld

Source: PCWorld

The behavior has been confirmed across Windows 11, Apple Silicon, and Ubuntu systems, with user reports indicating downloads without user consent have occurred for approximately a year . Hanff conducted controlled testing using a fresh Chrome profile on macOS, relying on the operating system's filesystem event logs to verify the activity. According to his analysis, the browser created the model directory and downloaded the full payload in the background during what appeared to be idle browsing time, completing the transfer in just over fourteen minutes

2

.

What the 4GB AI Model Powers and Why It's Hogging Computer Storage

The weights.bin file supports on-device AI features including writing assistance, scam detection, autofill, and suggestion capabilities

1

. Because Gemini Nano is designed to run locally rather than relying on cloud-based models, it requires training parameters stored directly on user devices. This approach offers privacy benefits by processing data locally, but creates significant storage challenges for users with limited disk space

1

.

Source: TechSpot

Source: TechSpot

Ironically, the 4GB folder isn't used for Chrome's most visible AI features. The AI Mode that appears in the address bar and Google Search runs on Google's servers, not from the locally stored weights

3

. Users who haven't actively enabled or searched for the supported features might have the model stored on their device without realizing it serves any function they actually use

3

.

Privacy Laws and Legal Concerns Raised by Researcher

Alexander Hanff, also known as "That Privacy Guy," argues that silently pushing a 4GB AI model likely violates provisions of EU law, including the ePrivacy Directive's rules on storing data on user devices and GDPR requirements around transparency and lawful processing

2

. While these claims haven't been tested in court, they reflect growing tension between aggressive feature rollout and regulatory expectations, particularly in Europe

2

.

Google does specify that "Gemini Nano's exact size may vary as the browser updates the model," but this information appears in a lengthy guide for built-in AI features rather than at the point of enabling them in Chrome

1

. The lack of clear notification about storage requirements or an option to power Chrome AI features with cloud-based models has created confusion and frustration among users

1

.

Bandwidth Consumption and Environmental Impact at Scale

Beyond storage space concerns, Hanff highlights the environmental impact of distributing a file of this size globally. If deployed across hundreds of millions or billions of devices, he estimates the total emissions impact of simply distributing the file could reach tens of thousands of tons of CO2 equivalent

2

. A "mid-band" deployment hitting 500 million devices would result in roughly 30,000 tonnes of CO2e, equivalent to the annual emissions of 6,500 cars

4

.

For users with metered or capped data connections, bandwidth consumption presents immediate financial concerns. A 4GB download is trivial on unlimited fiber connections, but for users on mobile hotspots, rural connections, or in developing markets where data is expensive, silently transferring gigabytes can have real financial consequences

2

.

How to Remove the OptGuideOnDeviceModel Folder

Users cannot simply delete the weights.bin file to free up storage space, as Chrome will re-download it if AI features remain enabled

1

. To prevent the file from returning, users need to navigate to Settings>System and toggle off the On-Device AI option

1

. Alternatively, users can type chrome://flags into the address bar, search for "Enables optimization guide on device on Android," and set the dropdown to Disabled

3

. This flag also indicates whether a device is eligible for the feature, as older hardware might not qualify

3

.

On Windows 11, the OptGuideOnDeviceModel folder resides at %LOCALAPPDATA%\Google\Chrome\User Data\OptGuideOnDeviceModel

3

. The most reliable way to remove it permanently is to uninstall Google Chrome entirely

3

. Multiple news outlets reached out to Google for comment but had not received responses at the time of publication

1

4

.

Today's Top Stories

TheOutpost.ai

Don’t drown in AI news. We cut through the noise - filtering, ranking and summarizing the most important AI news, breakthroughs and research daily. Spend less time searching for the latest in AI and get straight to action.

Instagram logo
LinkedIn logo
Youtube logo
© 2026 TheOutpost.AI All rights reserved