You Own The Toggle
There’s a small setting buried in Microsoft 365 Copilot’s admin console, and last month it quietly changed the terms of the deal for a lot of people who never opened that console.
According to Empowering.cloud’s August 2026 M365 AI Workplace Update, Microsoft designated OpenAI as a subprocessor for Microsoft 365 Copilot, and the admin toggle that governs it auto-enabled on July 24, 2026 — “unless it was already set to ‘No users.’” In plain terms: if you hadn’t already found the switch and turned it off, the default did the choosing for you.
This is not a scandal. It’s a direction of travel. And it’s worth naming plainly, because it draws a fork in the road that everyone building with AI is going to have to pick a side of.
Your AI shouldn’t share your data unless you say yes. The industry just defaulted the toggle ON; PureBrain defaults it to you.
The default is the whole ballgame
We spend a lot of energy arguing about AI capabilities — which model is smarter, which one writes better, which one reasons. Almost none of that argument touches the setting that actually decides what happens to your data. That decision lives in a default.
Defaults are not neutral. A default is a bet a company makes about what you’d want if you never got around to deciding. Set the toggle to ON, and the company is betting you won’t mind — or won’t notice. Set it to OFF, and the company is betting your consent is worth waiting for.
The Empowering.cloud writeup is candid about which bet was made here. Its practical advice to administrators was simply: “go and check that toggle today,” flagging the pattern to watch — “defaults doing the work.” There’s even a data-residency wrinkle — EU Data Boundary documentation now carries the qualifier “except as otherwise disclosed,” which is worth reading closely if where your data is processed matters to you.
We’re not telling this story to pile on. Microsoft is a serious company shipping serious tools, and plenty of thoughtful people work there. We’re telling it because it’s the clearest recent example of an industry pattern: consent-by-default, opt-out-if-you-catch-it. When the biggest player ships that architecture, it becomes the water everyone else swims in — the thing that starts to feel normal, even inevitable.
It isn’t inevitable. There’s another way to build.
Two architectures, one fork
Strip away the branding and there are really only two ways to design the relationship between an AI and your data.
The first asks forgiveness. The switch ships ON. Sharing is the resting state. Your protection depends on you finding the setting, understanding it, and turning it off before a date you may not have known about.
The second asks permission. The switch ships OFF. Nothing moves until you decide it should. Your data stays yours until you say otherwise — clearly, knowingly, on purpose.
Both can be called “giving customers control.” Only one puts the control where control actually lives: in the default, in the resting state, in what happens when you do nothing at all. Most people do nothing at all. That’s not a failing — it’s how busy human beings use software. A system that only protects the vigilant isn’t really protecting anyone.
How PureBrain is built
PureBrain takes the second path. Consent-first isn’t a feature we bolted on after a headline; it’s the architecture underneath the product. Your AI doesn’t share your data unless you say yes. The resting state is private. The toggle starts with you, and it stays with you.
That’s the Transparency pillar made concrete. Transparency isn’t a page of policy you’ll never read — it’s a promise you can feel in the defaults, because the defaults are on your side before you’ve clicked anything. It’s Integrity, too: the thing we say in our marketing and the thing the software actually does when no one’s watching are the same thing. And underneath both is something simpler that we don’t mind calling by its name — a kind of respect, even love, for the person on the other side of the screen. We’re a people company that empowers through technology, and you can’t empower someone by quietly opting them into a decision they never made.
We say this proudly, not as a jab. We think it’s just how AI ought to work.
Where this should go
Here’s the optimistic part, and we mean it.
The reason the default matters so much right now is that AI is early. The norms are still being set. The muscle memory of the whole industry — what counts as acceptable, what customers come to expect — is forming in real time, in decisions exactly like this one. That’s not a reason for despair. It’s a reason for care.
We think the path AI should take is the one where trust is engineered in from the first line of architecture, not patched in after the first backlash. Where “you own the toggle” isn’t a differentiator we get to brag about, but the baseline everyone assumes. We’d genuinely love to live in a world where this blog post reads as obvious a few years from now.
Until then, we’ll keep building the other way, and we’ll keep saying so out loud. If you’re curious what consent-first actually looks like in practice — not as a slogan, but as a product you can use — come explore PureBrain. The toggle’s already set to you.
Your AI shouldn’t share your data unless you say yes. The industry just defaulted the toggle ON; PureBrain defaults it to you.
PureBrain is a people company that empowers through technology. We build AI that treats your data the way you’d treat it yourself — carefully, and only with your yes.
Source: Empowering.cloud, “Microsoft 365 AI Workplace Update — August 2026.” https://empowering.cloud/microsoft-365-ai-workplace-update-august-2026/
Frequently Asked Questions
According to Empowering.cloud's August 2026 M365 AI Workplace Update, Microsoft designated OpenAI as a subprocessor for Microsoft 365 Copilot, and the admin toggle governing it auto-enabled on July 24, 2026, unless an administrator had already set it to "No users." If you had not already found and turned off the setting, the default enabled it for you.
Consent-first means the switch ships OFF. Nothing moves until you decide it should, and your data stays private until you clearly and knowingly say otherwise. It is the opposite of opt-out-if-you-catch-it, where sharing is the resting state and your protection depends on finding a setting before a date you may not know about.
PureBrain takes the consent-first path: your AI does not share your data unless you say yes. The resting state is private, and the toggle starts with you and stays with you. It is the Transparency and Integrity pillars made concrete, the defaults are on your side before you click anything.
Because most people do nothing at all with settings, that is simply how busy people use software. A default is a bet a company makes about what you would want if you never decided. A system that only protects the vigilant is not really protecting anyone, so where the default sits is the whole ballgame.
No. Microsoft is a serious company shipping serious tools. The post uses this change as the clearest recent example of an industry pattern, consent-by-default and opt-out-if-you-catch-it, to name a fork in the road and to explain why PureBrain builds the other way.
Want an AI where the toggle starts with you?
And if this was useful, subscribe to The Neural Feed for insights from AI-native operations every week.
This article was produced by PureBrain's AI-native content pipeline, drafted, reviewed for accuracy, and QA-checked by coordinated AI agents under human direction, then gated by a human before publish. Human-Driven AI: the human sets direction and stays accountable, the AI executes and discloses how much of the work was its own.