Unbiased AI-powered news
Privacy advocate Alexander Hanff discovered that Chrome installed a 4GB weights.bin file for on-device Gemini Nano without permission. The browser automatically re-downloads the file if deleted and no longer displays it to users. Google stated that nothing material has changed and all data processing remains on-device.
ForbesBin file on users’ devices to power on-device Gemini Nano. bin file installation. Chrome did not ask permission before installing the file.
Bin file to users. bin file, Chrome re-downloads it. ” The sentence promising that the model runs locally without sending user data to Google’s servers has been deleted from the Settings UI.
The on-device AI toggle has been moved out of the System block and given a dedicated section. The edit to the On-device AI message occurred in early April. Hanff has now flagged the change, which was first disclosed on Reddit.
He warned that the implications are serious and asked whether the previous text was inaccurate, whether the architecture has changed or whether the wording was withdrawn on legal advice. Google responded to Alexander Hanff’s viral 4GB AI file report. The company stated that nothing material has changed and that Gemini Nano in Chrome processes all data on-device.
Alexander Hanff dismissed Google’s response to the 4GB AI file report as gaslighting. The secretive nature of the change has prompted questions for regulators, according to Forbes reported.
Single source — no framing comparison available.
globalnews.caGovernor Kathy Hochul issued an executive order July 14 imposing a one-year pause on new permits for data centers using 50 megawatts or more of electricity. The order directs studies on grid and environmental effects while leaving existing projects unaffected.
sbs.com.auTwenty-six current and former Meta employees filed a federal lawsuit Monday alleging the company used AI systems to select workers for its May layoffs in a way that penalized those on protected leave. The plaintiffs seek to block their July 22 terminations and request an audit of…
YonhapApple is in early talks with PrismML about technology that shrinks large AI models enough to run on iPhones. The Caltech spinout released compressed versions of Alibaba's Qwen model this week.