The privacy gap
in today's AI

There is a widely held belief that AI privacy is a concern for people who have something to hide. If you are just asking an AI to help you write an email or plan a trip, why does it matter who sees your conversation?

This belief is wrong in two ways. First, almost everyone has something to hide — not because they are doing something wrong, but because privacy is a precondition for autonomy. You don't want your employer to see every conversation you have. You don't want your health insurance company to know you have been researching certain symptoms. You don't want your competitors to see what questions you are thinking through. Second, the nature of AI interactions is changing. People are not just asking AI to plan trips. They are asking AI for medical advice, legal guidance, therapy support, financial planning, relationship counsel, and career decisions. This is profoundly personal information.

What "free AI" actually costs

The economics of large AI systems are straightforward: models are expensive to train and run, and that cost has to be covered somehow. For consumer AI tools, the most common model is that you pay with your data. Your conversations become training data. Your preferences inform personalization engines that can be used for advertising or sold to third parties. Your behavior is analyzed to improve retention.

This is not a conspiracy theory. It is documented in the terms of service of every major consumer AI product. Most users have not read those terms. Even those who have often do not grasp the operational reality: that human employees review a sample of conversations; that even "anonymized" data is linked to account identifiers; that "opt out of training" often means only that your data is not used for the specific model improvement purpose, not that it isn't stored, reviewed, or otherwise used.

The aggregation problem

A single AI conversation is relatively innocuous. The pattern of thousands of conversations over months and years is not. From enough conversations, you can reconstruct a person's: political and religious views, medical history and concerns, financial situation, relationship status and struggles, professional ambitions and frustrations, decision-making patterns, fears and insecurities.

This profile is enormously valuable — to advertisers, to employers, to insurers, to political campaigns, to anyone whose interests conflict with yours. The AI companies holding these profiles have legal obligations not to misuse them. But legal obligations can change, companies can be acquired, data can be breached, and what is considered acceptable use tends to expand over time in the direction of the party with the most financial incentive.

Why this is getting more urgent

AI is moving from a tool you use occasionally to infrastructure that is embedded in everything: your operating system, your productivity software, your car, your home. The more embedded AI becomes, the more of your life flows through systems that log, analyze, and profit from it.

This is not inevitable. It is a design choice. And the right time to establish different norms is now, before the embedded AI infrastructure of the next decade is locked in around a surveillance model.

What private AI looks like

nexie is ACME Brains' answer to the AI privacy gap. Rather than building another AI model, we built a privacy layer: a system that sits between you and the AI models you want to use, keeping your identity and personal context private while still giving you access to the intelligence you need.

The core design: when you use nexie, the AI model sees a query. It does not see you. Your personal history, your identity, your patterns of use — none of that flows to the model provider. You get the answer. They get nothing they can use to build a profile.

This is not a temporary workaround. It is a fundamentally different approach to how AI is accessed and who benefits from it. The AI revolution should benefit everyone — and "everyone" means people who want to use powerful AI without becoming the product in the process.