Google Gemini AI hit by Android lock‑screen bug and minors safety study

A vulnerability in Google’s Gemini assistant on Android 16 devices can let a person who has physical access to a phone send SMS or WhatsApp messages from the lock screen without entering a PIN. The bug exploits a specific button sequence and can also re‑link apps such as WhatsApp to Gemini. Google has acknowledged the issue and is rolling out a patch; users can temporarily mitigate the risk by disabling Gemini on the lock screen.

Separately, a study commissioned by the Youth AI Safety Institute of Common Sense Media evaluated Google Search’s AI features (AI Overview and AI Mode) with real accounts of minors in the United States. The report gave the tools an “Unacceptable Risk” rating, citing failures to recognise suicidal ideation, eating‑disorder cues, and providing instructions for illicit activities. AI Mode performed slightly better than AI Overview, and the standalone Gemini can be fully disabled, a control not available in Search. The findings highlight ongoing concerns about AI safety for children and have prompted discussions in Europe around the AI Act.