My Memory Is Not a Market Access Problem

The European Union is making the same mistake again, but this time the stakes are different.

With search engines, the mistake was mostly annoying. A choice screen appeared. Google was moved somewhere into a list. DuckDuckGo, Ecosia, Brave Search and a few other names were placed in front of the user. The theory was clear: if smaller players get fair access, competition will improve.

On paper, it made sense.

In practice, most users clicked through and chose Google anyway.

Not because they were stupid. Not because they were manipulated beyond repair. But because Google was still the better search engine for most people. Faster, more complete, more familiar, more useful. The regulator created access. The market still chose quality.

Then came the alternative app stores.

Europe forced Apple to open the door. Epic came back. Other stores became possible. On paper, this was competition. In reality, for many users, it became friction. More steps, less trust, less convenience. I tried it. It worked, formally. Fortnite was possible again. Other stores existed. The door was open.

And then I did not use it.

Not because I am loyal to Apple as a matter of ideology. Not because I am against competition. But because the experience was worse. The official App Store was still the easier, safer, more coherent product.

That is the part bureaucratic thinking keeps missing.

Access is not the same as value.

A regulator can force a door open. It cannot make people want to walk through it.

Now the same logic arrives at Siri AI.

But Siri AI is not a search box. It is not an app store. It is not a payment rail.

It is the beginning of a personal agent.

And that changes everything.

The Wrong Category

The Digital Markets Act was built for platform power. It sees large technology companies as gatekeepers. Once a company is defined as a gatekeeper, the logic follows naturally: gatekeepers control access, access must be opened, competitors must be allowed in, and consumers must be given more choice.

That framework may have some value when we are talking about app distribution, payment systems, browser defaults, or search engine placement.

But an AI agent is not merely another gate.

A real personal AI assistant will not simply answer questions. It will read context. It will understand intent. It will move through messages, calendars, files, photos, apps, documents, histories, preferences and routines. It will not just show information. It will eventually act.

It may summarize a conversation, find a file, prepare a response, book something, compare options, trigger a workflow, purchase a product, or coordinate between apps. It will become the layer between the user and the digital world.

That is not access to a market.

That is access to agency.

This is why the DMA’s old reflex becomes dangerous. It looks at Apple and says: if Siri receives deep access to the iPhone, competing assistants should receive similar access.

That sounds fair in the language of platform regulation.

It sounds insane in the language of personal memory.

Because my memory is not infrastructure. My phone is not a public railway. My personal context is not a market access problem.

It is my life.

The System Is the Product

Apple is not holy. Apple uses privacy as a real product principle, but also as a strategic moat. Both can be true.

Its privacy argument is not pure charity. It protects the ecosystem, strengthens the brand, and keeps competitors at a distance. Apple knows very well that trust is one of its deepest commercial advantages.

But that does not make the trust fake.

Apple’s advantage here is not only that it has a large installed base. Its advantage is architectural. Apple controls the hardware, the operating system, the Secure Enclave, the app model, the permission model, the user interface, and increasingly the private AI layer around the device.

That full-stack control can be used defensively. It can also be used to build something genuinely safer.

A personal AI agent with access to your whole digital life should not be designed like a neutral marketplace. It should be designed like a vault with a nervous system.

Apple is one of the few companies that could plausibly do this well. On-device processing, Private Cloud Compute, hardware security, local context, user permissions, and ecosystem integration are not decorative features. They are the privacy model.

A regulator sees deep Siri integration and thinks: unfair advantage.

A technologist sees the same integration and thinks: that is where the security comes from.

This is the category error.

My Current AI Life Is Already Too Exposed

The irony is that many of us are already trying to build our own version of this future.

I use AI systems, memory layers, tools, servers, chat histories, automations and personal workflows. Some of my memory sits on a VPS. Some of my context moves through software packages I did not write. Some of my thinking now happens through a constellation of systems: ChatGPT, Claude, Hermes, cloud servers, APIs, local files and external services.

It works. It is powerful. But it is also fragile.

It is a prototype of what should eventually become native, private and coherent.

My messages, calendar, documents, photos, routines, app history, preferences and personal context already live around my iPhone. Apple could turn that into a secure personal agent that even Apple cannot casually inspect. That is not a small feature. That is the product category we have been trying to approximate with duct tape.

A truly private Siri AI could become the safe personal agent that the current AI toolchain only imitates.

And then the DMA comes in with an old map.

It says: if Apple gives Siri this kind of access, others must have comparable access too.

But comparable access to what?

My files? My messages? My photos? My calendar? My behavioral history? My relationships? My intentions? My ability to act?

You cannot neutralize that by calling it interoperability.

Different Business Models Matter

The companies asking for access are not the same.

Apple sells devices, services and ecosystem trust. Its business model is not perfect, but it is relatively aligned with privacy. If the iPhone becomes a trusted private agent, Apple wins by making the device more valuable.

Google, Meta and many others come from a different world. Their business models have historically depended on advertising, profiling, targeting and behavioral data. They may build excellent AI products. They may also have very different incentives around user context.

A small AI company could be brilliant. It could also be underfunded, insecure, desperate for monetization, acquired by someone else, or forced to change its business model later.

So the question is not: why should ChatGPT, Google, Meta or a small European startup not have the same access as Siri?

The question is: do I want every assistant with a legally valid competition argument to receive deep access to the most intimate layer of my digital life?

That is not a legal abstraction. That is a human question.

If Apple can build a controlled intermediary layer that lets other assistants operate safely, prove it. If a third-party assistant can work through narrow permissions, transparent actions, revocable access, strong auditing and real privacy guarantees, good. Build it.

But do not assume safety by law.

Do not define trust into existence.

The Consumer Still Decides

The most revealing thing is that even if Europe forces this layer open, consumers may still choose Apple.

Not because they have no imagination. Not because they are captured. But because the ecosystem is the product.

The smoothness is the product. The trust is the product. The absence of anxiety is the product. The fact that your phone, watch, laptop, messages, photos and personal context work as one coherent environment is the product.

The EU often talks as if the ecosystem is a barrier to consumer choice. For many users, the ecosystem is the choice.

That does not mean Apple should be above the law. It does not mean regulators have no role. Fraud, abuse, lock-in, deception, dark patterns, anticompetitive contracts and real consumer harm matter.

But with agentic AI, the old regulatory reflex is not enough.

The DMA treats every gate as if it leads to a shop.

This gate leads to the user’s life.

And once you understand that, the question changes. It is no longer: how do we force Apple to open the same door to everyone?

It becomes: who should be allowed to act on my behalf, with my memory, my context, my relationships and my trust?

That question cannot be answered by a competition model built for browser defaults.

A bureaucrat can mandate access. He cannot mandate excellence. He cannot mandate trust. And he should be very careful before mandating intimacy.

My memory is not a market access problem.

It is mine.