What to like about Personal Intelligence
If Gemini helps you keep up with life and work, giving it access to what Google already knows about you makes some sense.
Once you turn it on, Personal Intelligence can tap into the contents of your Gmail, Google Photos, past Search queries and YouTube history when you ask it a question. That means it can use your data to inform its “reasoning,” better understanding your personal context and surfacing personal insights in its answers without you having to tell it where to look.
For example, after switching on Personal Intelligence, I tried asking: “When should I make my next appointment for a haircut?” After thinking for a few seconds, Gemini found records of my older past appointments in my Gmail and, based on that information, told me I should make another appointment for next week. It doesn’t quite rise to the level of a personal assistant, but certainly a competent intern. (Apple’s Siri can’t answer a question like that.)
Personal Intelligence is also being integrated into the AI Mode that’s part of Google search, so it can use your preferences or existing plans to inform your hunts for information and products.
Google says Personal Intelligence is available to personal Gemini users with paid accounts in the US, and it plans to expand over time to free users and more countries.
What to worry about
Google says turning on Personal Intelligence doesn’t give it the right to train its AI directly on your Gmail inbox or Google Photos library. There are no targeted ads in Gemini, for now.
But there’s a different kind of privacy problem to consider: giving Google permission to mix together information from different contexts in your life could lead to embarrassment, or worse.
The issue is called “data bleed,” says Miranda Bogen, director of the AI Governance Lab at the Center for Democracy and Technology. Everybody has personal information that’s appropriate in certain conversations that doesn’t belong in others. Think: medical information, emotional support from a friend and data going into a work report.
Once Gemini brings a piece of personal context into your chat history, it could pop back up at an inappropriate time. “Now it’s all getting put into one big blob and moved back and forth across different products,” Bogen said. “It is harder and harder to keep track of what’s even happening, let alone draw lines around when something shouldn’t happen.”
For example, she said information regarding your health records sitting in your email could be pulled into a chat while you are drafting a message to your boss. Or, the AI might rely on proprietary work documents when answering a personal query about your retirement portfolio.
The AI could also misunderstand you, and its prediction of what is “most useful” may not actually align with your best interests. Or it could pigeonhole you.
Google acknowledged in a white paper it released with Personal Intelligence that Gemini currently struggles to understand context perfectly. For example, it said if Gemini spots an email about your employment, it might start anchoring all its responses around the fact that you’re a software engineer. Or, based on a receipt in your email, Gemini might keep suggesting heavy metal concerts nearby, when you actually purchased the tickets as a birthday gift for your brother.
Google says its latest Gemini 3 has an improved grasp of nuance, allowing it to better interpret personal context and determine exactly when to retrieve relevant information. And it has built some guardrails: Gemini is designed to avoid making proactive assumptions about sensitive data such as your health, although it will discuss that data if the user explicitly asks.
Google also says its primary defence against unwanted data mingling is user choice. Personal Intelligence is off by default, though I noticed Google kept nudging me to turn it on when I ignored it at first. You can also choose which data sources Gemini can access – for example, you could connect Gmail but not Google Photos.
Once Personal Intelligence is on, you can also tell Gemini to not personalise a specific response – but only after the fact. You have to tap a button and tell it to answer again without using your data.
What you should do
To mitigate these risks, Bogen recommends thinking carefully about what information such as health data exists in your email and photos, and whether you are comfortable with an AI potentially accessing it.
And treat it as a sandbox, for now. Turn on Personal Intelligence to experiment and see how it works, but then turn it off rather than leaving it permanently enabled.
“The whole field is moving so quickly … a choice you think you’re making today might actually have different implications tomorrow,” she said.
Sign up to Herald Premium Editor’s Picks, delivered straight to your inbox every Friday. Editor-in-Chief Murray Kirkness picks the week’s best features, interviews and investigations. Sign up for Herald Premium here.
