You just got a new Chrome update. Routine stuff. Nothing to worry about.
Except buried inside that update is a Google AI assistant called Gemini — already running in your employees' browsers, already enabled, already reading whatever web page happens to be open. No one asked you. No one asked your employees. It just showed up.
Google's position is essentially: trust us, we handle your data responsibly. Which is a little like a locksmith installing a new door on your building, handing themselves a key, and telling you not to worry because they promise not to snoop around.
Welcome to the Barn Door Problem.
What Exactly Is "Gemini in Chrome"?
First, let's clear something up — "Gemini" isn't one thing. It's actually three distinct products that Google has quietly rolled into your browser and productivity suite, each with a different risk profile.
Gemini Nano is the least concerning of the three. It's a compressed AI model that Chrome downloads and runs entirely on the local device. Nothing gets sent to Google for processing — the AI inference happens on your PC's GPU or CPU. Think of it as a small AI brain living inside the machine itself. The catch: it only runs on newer hardware with specific processor requirements, so most business PCs don't even support it. And even though inference is local, Chrome still phones home to Google for model updates and usage telemetry.
The Gemini Chrome Sidebar is where IT managers need to pay attention. This is a browser panel that can read the active web page and respond to employee prompts. The key word there is read. Whatever page is open — your ERP system, HR portal, customer database, internal pricing tool — the sidebar can ingest that content and send it to Google's cloud servers for processing. It ships enabled by default. No administrator decision required. It's just on.
Gemini for Google Workspace is the deepest integration. This is the AI baked into Gmail, Docs, Drive, and Meet. It has broad access to your entire organizational data corpus — every email, every document, every shared file. It operates under your Google Workspace data processing agreement, with one important catch: most organizations haven't verified whether their existing agreement actually covers the AI features. Many older Workspace contracts were signed before Gemini existed.
The Scenario Nobody Thinks About Until It Happens
Your accounts payable coordinator opens Chrome. She's got four tabs open: Gmail, your ERP billing module, a vendor contract in Google Docs, and the Gemini sidebar, which she started using last week because it helps her draft emails faster.
She asks Gemini to help summarize a vendor dispute. Gemini helpfully reads the context from her open tabs — including the ERP screen showing invoice amounts, the contract terms in Docs, and the email thread she's working on.
All of that just went to Google's cloud.
Did she do anything wrong? Not intentionally. Did it violate your data policies? Almost certainly. Did anyone tell her this was happening? No — because nobody told IT it was happening either.
That's the barn door. It wasn't kicked open by someone breaking in. Google just built it that way and left it unlocked.
Who This Hits Hardest
This isn't just an abstract data hygiene concern. Certain industries and environments face real compliance exposure:
Manufacturing companies handling controlled technical data, customer proprietary specifications, or defense-adjacent work need to know that the Gemini sidebar doesn't distinguish between a public webpage and a proprietary engineering document. It reads what's open.
Construction companies managing bid data, subcontractor financials, and owner contracts have competitive information flowing through their Workspace environments every day. Gemini for Workspace has a front-row seat to all of it.
Any company with a Google Workspace contract needs to pull that agreement out and actually read the AI data handling provisions. Spoiler: if you signed it before 2023, Gemini probably isn't in there.
Three Things to Do This Week
The good news: you have granular control over this. Google built Chrome Enterprise policy controls for each Gemini feature individually. Here's what to prioritize:
1. Push a master kill switch via GPO or Intune.Set the Chrome Enterprise policy GenAiDefaultSettings to 2. This disables all experimental AI browser features across your managed devices in one shot. Add SideSearchEnabled = false as belt-and-suspenders specifically for the sidebar. This takes about 15 minutes to configure and deploy.
2. Verify it actually worked.Open chrome://policy on a test machine. Every policy you pushed should show status OK. If something shows Error or is missing, it is not enforced — regardless of what your GPO console says. Screenshot it and keep it on file.
3. Check your Google Workspace contract.Find your Data Processing Agreement (or Addendum) with Google and verify it explicitly covers Gemini AI features. If it predates 2023, call your Google account rep and ask for an updated DPA. While you're at it, turn off the "Help improve Gemini apps" toggle in Google Admin Console — that's the setting that allows employee prompts to be used for model training.
The Right Way to Think About This
Disabling Gemini org-wide isn't a permanent answer — AI tools genuinely have productivity value, and your employees will find workarounds if you just say no without a plan.
The right posture is: disable by default, authorize by exception.
Turn it off across the board first. Then work through which use cases actually make sense, which employees handle data that rules certain features out, and what your Google contract actually permits. Re-enable selectively for approved groups, with controls in place.
That way you decide who gets the key to the barn — not Google.
Need Help Sorting This Out?
Infortech works with manufacturing and construction companies throughout Silicon Valley to get their AI governance house in order before something goes sideways. We can review your Chrome Enterprise policy configuration, audit your Google Workspace data agreements, and help you build an AI acceptable use policy that actually maps to how your business works.