What Actually Triggered the Decision
It was not a data breach. No account compromise. No news story about a privacy scandal, no leaked internal document, no third-party researcher publishing something alarming. What changed his mind was reading the actual terms of service , specifically the clauses covering how Google Workspace data is used for model training.
He had agreed to those terms years earlier without reading them carefully. Most people have. When he finally did read them, line by line, with the intent to actually understand what he had agreed to, he decided the trade-off was not one he wanted to make any longer.
That distinction matters a great deal for understanding what this story is actually about. This is not a piece about Google doing something wrong or secretly harmful. It is about someone making a deliberate choice after understanding what an agreement he had already signed actually said. Two different things.
What Google Actually Collects and What It Does With It
Every search query. Every document opened in Workspace. Every email processed through Gmail. Every conversation with Gemini. Every file stored in Drive and accessed through Google's interface. The terms of service permit Google to use all of this data to train and improve future models.
This is not hidden in fine print designed to obscure it. It is in the terms, it has been discussed in Google's public communications, and Google has not been deceptive about the general shape of it. The question is not whether the collection happens , it does , but whether most users understand what they agreed to when they signed up for productivity tools that are free at the point of use.
The deeper structural issue is integration. Google's AI features are not optional add-ons sitting alongside the core tools. They are built into Gmail, Docs, and Drive at a level where opting out of AI features often means opting out of meaningful functionality in the tools themselves. The productivity software and the data collection are not cleanly separable.
This makes the opt-out calculation harder than it appears. You are not just deciding whether you want Gemini writing your emails. You are deciding whether you want to keep using the integrated suite at all, because the suite is designed to make the AI features feel like the natural way to use it.
The Specific Replacements He Made
Search moved to Kagi, a paid search engine with no advertising model and explicit policies about not using search queries for model training. The lack of an ad model changes the incentive structure in a meaningful way , Kagi is paid directly by users, which means the business does not depend on maximising engagement or data collection.
Email moved to Proton, which uses end-to-end encryption by default and is headquartered in Switzerland under legal frameworks that differ from those governing US-based providers in significant ways. End-to-end encryption means Proton cannot read the content of emails even if compelled to, because they do not hold the keys.
Documents and notes moved to Obsidian with local files and no cloud sync enabled by default. No third-party server is processing the content. The files exist on his own hardware. This trades the convenience of cloud access from any device for the assurance that the content is not being processed anywhere he has not explicitly chosen.
For AI assistance, he split his approach by what the conversation contained. Local models via Ollama for anything he considered sensitive or private , conversations that stay entirely on his own hardware. Claude Projects for everything else where the data sensitivity felt lower and the quality of the assistance mattered more than complete data isolation. He did not treat all AI assistants as equivalent risks and made conscious distinctions based on content.
The Honest Cost of Leaving Google
The financial cost was approximately $40 per month. Kagi, Proton, and the other privacy-focused tools are not free services. This is worth being direct about: the "free" Google suite was never actually free in a meaningful sense , the price was data and the implicit agreement to its use , but the cash cost of the alternative is real and not everyone can or will absorb it. $40 per month is not nothing.
The productivity cost was harder to quantify with precision, but he estimated roughly 15% more friction in daily workflows. Some integrations that worked smoothly inside Google's ecosystem do not have equivalents across the alternative tools. Calendar connections, mobile sync on certain platforms, collaboration with colleagues and clients who are still on Google , all of it requires workarounds that add steps and take time.
There are also capability gaps. Google's AI features, when they work as intended, genuinely accelerate certain tasks. Gemini summarising a long email thread saves real time. Smart suggestions in Docs reduce certain kinds of editing friction. The alternatives do not replicate all of this. Some of the functionality is simply gone, not replaced.
He did not pretend these costs were trivial or that the transition was painless. The point of the exercise was not to claim the alternative stack is better in every measurable way. It is to make the choice with accurate information about what each option actually involves, rather than defaulting to convenience and never examining the terms at all.
The Honest Case for Staying on Google
Google's AI features are genuinely useful. Gemini integrated into Docs, Gmail, and Drive can save meaningful time on real tasks for real users. The integration across tools is seamless in a way the alternatives have not come close to replicating. The mobile experience is strong. The collaboration features work well with people who are also on Google, which is most people.
He acknowledges this directly. The case for staying on Google is not a weak one and he does not pretend otherwise. For the vast majority of users in the vast majority of use cases, the practical privacy exposure is low. The data is used for training models in aggregate , not targeted at you individually, not used to serve you specific ads based on document content, not handed to advertisers in identifiable form.
The concerns are real but they are theoretical in the sense that most users will never encounter a concrete harm from the data collection. The information goes into training datasets. Future models become somewhat better because of it. Your specific emails or documents are not being read by anyone at Google with an interest in your particular situation.
What changed for him was not the risk level in absolute terms. It was his preference for where his data lived and who processed it, once he understood exactly what the default arrangement was. That is a personal judgment, not a universal conclusion about what everyone should do.
What the Recommendation Actually Is
The recommendation at the end of this story is not that you should de-Google your life. The recommendation is narrower and more honest than that: you should make whatever decision you make with full information about what the trade-off is, rather than continuing by default without having examined it.
Most people using Google's entire productivity suite have not read the terms that govern how their data is processed. That is normal. The terms are long, written in legal language, and updated periodically without much notice. Reading them in full is not a reasonable expectation for most users. But the information is there, and someone who wants to understand what they have agreed to can do so.
Some people will read those terms and decide the trade-off is fine , the tools are useful, the practical risk is low, and the $40 per month alternative is not worth it. That is a completely defensible position. Some people will read the same terms and decide, as he did, that the arrangement is not one they want to continue with. That is also a defensible position.
Free tools are rarely free in every sense.
Sometimes the cost is money you would have paid anyway.
Sometimes it is data, and that is worth knowing before you decide whether the exchange is one you want to continue.