Chrome removes claim that its 'on-device AI' does not send data to Google servers.
Google Chrome downloads '
Chrome removes claim of On-device Al not sending data to Google Servers. : chrome
https://old.reddit.com/r/chrome/comments/1t5qayz/chrome_removes_claim_of_ondevice_al_not_sending/
Chrome Deleted Its Own Privacy Promise for Sneaky On-Device AI - Decrypt
https://decrypt.co/367193/chrome-removes-privacy-claim-gemini-nano-google
'On-device AI' is a system that runs AI models directly on the user's PC or smartphone, rather than sending data to a cloud server for processing. Google has integrated a lightweight AI model called 'Gemini Nano' into Chrome, enabling it to process things like fraudulent site detection, web page summarization, and AI features for developers in a local environment. Chrome 148, released on May 5, 2026, also introduced the 'Prompt API,' which allows websites to directly access the browser's built-in on-device AI. However, privacy audit expert Alexander Hanff points out that 'Google Chrome downloads approximately 4GB of on-device AI models to the user's PC without obtaining clear confirmation.'
Google Chrome is being accused of secretly saving approximately 4GB of on-device AI models; even after deletion, they are sometimes re-downloaded - GIGAZINE
The fact that the privacy wording regarding this on-device AI had been quietly changed became a topic of discussion in the Chrome community on Reddit.
The following is a screenshot of Chrome 147 posted to the community. The section indicated by the blue line contains an explanation of On-Device AI, stating that it 'works without sending data to Google servers.'
However, in the following screenshot after updating to Chrome 148, the phrase 'Do not send data to Google servers' has disappeared from the explanation regarding on-device AI.
A Reddit post expressed concern that 'the removal of the explanation means that data will now be sent to Google.'
This change was picked up by the social news site Hacker News and became a hot topic. Criticisms included 'This is basically AI spyware installed on your device,' and sarcastic comments on Google's methods such as 'I think embedding AI in desktop apps and sending data to headquarters for processing is a great way to collect data from people who will never even notice.' There were also comments criticizing Google, such as 'I'm surprised there are still people using Chrome when Google has done untrustworthy things so many times and tried to exploit users whenever it had the chance.' Another user pointed out that 'In my opinion, the essence of the AI business is data collection. The value of AI is not in the quality of the models that corporate clients and developers pay for, but in the amount of data that the host of the model can obtain for 'free.' And the value of that data is determined by how much value the buyers, such as insurance companies and advertisers, see in it.'
Regarding the removed explanatory text, a Google spokesperson told Decrypt, 'The removal of this phrase does not reflect any changes in how Chrome processes on-device AI. The data passed to the model is processed only on the device,' denying concerns that data would now be collected.
According to Google, websites using the on-device LLM 'Gemini Nano' in Chrome may allow users to see the input and output of the AI model. In such cases, the use of the data is subject to the website's individual privacy policy, and a spokesperson said they removed any explicit references to Google servers to avoid potential confusion.
Related Posts:
