Google’s Gemini chatbot can now remember things like info about your life, work, and personal preferences.
As flagged by posters on X (and Google’s official account), a “memory” feature has begun rolling out to certain Gemini users, including this reporter. Like ChatGPT’s memory, Gemini’s adds context to the current conversation. For example, tell Gemini to remember foods you like and the next time you ask the bot for restaurant recommendations, it might tailor its suggestions to your culinary leanings.
Memory is only available for subscribers to Google’s $20-per-month Google One AI Premium plan, per 9to5Google, and it hasn’t yet made its way to the Gemini apps on iOS and Android — just the web client.
In the Gemini UI, Google gives a few examples of potentially useful priming memories, like “Use simple language and avoid jargon,” “I can only write code in JavaScript,” and “When trip planning, include the cost per day.” The company notes that the feature, which only supports English-language prompts for now, can be switched off at any time — but that memories are stored until manually deleted.”
But Gemini memories aren’t used for model training. “Your saved information is never shared or used to train the model,” a Google spokesperson told TechCrunch via email.
Memory features such as ChatGPT’s and Gemini’s can be exploited if not carefully engineered with guardrails. Earlier this year, a security researcher found that hackers could surreptitiously plant “false” memories in ChatGPT to effectively steal a user’s data in perpetuity.