Google has launched Personal Intelligence in the Gemini app for Zimbabwe users, a feature that lets the chatbot draw on a user’s Gmail, Google Photos, YouTube and Search history to produce personalised responses.
The feature, which rolled out in the US earlier this year, is being made available first to paying subscribers on Google’s AI Plus, Pro and Ultra tiers, with free users to follow “in the coming weeks”, according to the company. It works across web, Android and iOS, and across all models in the Gemini model picker.
Personal Intelligence is Google’s attempt to differentiate Gemini from rivals such as OpenAI’s ChatGPT and Anthropic’s Claude by leveraging data users have already stored in Google services. The company’s pitch is that because this information “already lives at Google securely”, users do not need to send personal data to a third party to get personalised AI responses.
Connecting apps is off by default. Users choose which services to link, and can disconnect them or delete chat history at any time, Google said.
The company said Gemini does not train directly on users’ Gmail inboxes or Google Photos libraries. Instead, it said, training is done on “limited info, like specific prompts in Gemini and the model’s responses”, with steps taken to “filter or obfuscate personal data” from conversations before they are used.
In an example used in a Google blog post by Josh Woodward, vice-president of the Gemini app, the company described a user asking Gemini for tyre recommendations. Gemini pulled the vehicle’s licence plate from a photo and its trim details from an e-mail to locate the right tyre size. Google said the photo and e-mail were “referenced to deliver the reply” but not used to train the model.
How those assurances hold up under closer scrutiny – and how they align with South Africa’s Protection of Personal Information Act – is likely to be a question for regulators and privacy advocates as the feature rolls out.
Google acknowledged the system can produce inaccurate responses or “over-personalisation”, where Gemini draws unwarranted connections between unrelated topics, and asked users to flag problems via a thumbs-down feedback mechanism.
