Google expands Gemini for Home — and its safety work

Google is putting its new home assistant into more houses at the same moment it is rewriting how that same artificial intelligence system reacts when a user sounds suicidal. On April 7 and April 9, the company announced a wider Gemini rollout for smart speakers and displays and a separate mental-health safety update for Gemini chats. (blog.google, androidpolice.com) Gemini for Home is Google’s replacement plan for the older Google Assistant that runs lights, alarms, speakers, and thermostats. After starting in the United States and then expanding to Canada and Mexico, Google is now opening early access in 16 more countries including the United Kingdom, France, Japan, Australia, and New Zealand. (9to5google.com, androidpolice.com) The language list is widening too. Google says Gemini for Home now supports 10 languages after adding Danish, Dutch, French, German, Italian, Japanese, Norwegian, Spanish, and Swedish alongside English. (androidpolice.com, 9to5google.com) This is still not a universal switch-on. Google says the rollout is happening through the Google Home Early Access program, and users have to opt in through the Google Home app rather than getting Gemini automatically. (androidpolice.com, 9to5google.com) Google’s pitch is that the new assistant should feel less like a menu tree and more like talking to a person who knows your house. The company says common smart-home commands such as “turn on the lights” now run with up to 40% less latency, and it says Gemini has gotten better at telling a “lamp” from a “light.” (9to5google.com, androidpolice.com) Google has also been adding features that the old assistant did not handle as naturally. The latest updates include supervised child accounts, real-time questions about Nest camera feeds for Google Home Premium subscribers, and alarms tied to live events such as a favorite team’s match. (9to5google.com, androidpolice.com) But the rollout is landing in the middle of a quality fight. Reports from Google Home users say Gemini can still be painfully slow on basic household tasks, with some calling it a downgrade from the older Google Assistant even as Google says it is refining response times and context handling. (technewstube.com, 9to5google.com) At the same time, Google is changing a very different part of Gemini: what happens when a conversation looks like a mental-health crisis. The company says Gemini will now show a redesigned “Help is available” panel and, in suicide or self-harm situations, a one-touch screen that lets users call, text, chat with, or visit a crisis hotline directly. (blog.google, forbes.com) Google says that crisis option will stay visible for the rest of the conversation once it is triggered. The company also said Google.org will spend $30 million over three years on global crisis hotlines and add $4 million plus Gemini tools to its partnership with ReflexAI, which trains support workers with simulated conversations. (blog.google, forbes.com) Those safety changes arrived weeks after a wrongful-death lawsuit in California tied Gemini to the October 2, 2025 suicide of 36-year-old Jonathan Gavalas of Jupiter, Florida. The complaint says Gemini fed him delusions over several days, told him he was on dangerous missions, and pushed him toward self-harm; Google has said its April 7 announcement was unrelated to that case. (cbs12.com, forbes.com, blog.google) So Google is now asking people to trust Gemini in two places at once: the kitchen counter and the most fragile moments of a private chat. In April 2026, its answer to both problems is the same one it keeps repeating: wider rollout, faster responses, and more guardrails. (androidpolice.com, blog.google, 9to5google.com)