Freedom in the Digital: Three Things That Have to Arrive Together

Sovereignty, reliable rules and real trust. Take one away and the other two stop working.

Tonight in Berlin, the Deutsche Gesellschaft für Recht und Informatik e.V. (DGRI) opened its annual conference, in its fiftieth year. The theme is a single question, and the society puts it plainly: Freiheit im Digitalen. How free are we in an increasingly digitised world? 🕊️

It is not a rhetorical question. The programme runs at the Humboldt-Universität in Berlin until 13 November, organised around the themes that take the question apart: artificial intelligence, digital sovereignty, anonymity, and freedom of expression. Judges, academics, regulators, platform operators and practising lawyers in the same room, which is rarer than it should be.

Our Managing Director is attending. Here is why the framing matters to anyone building AI systems for European organisations.

🧭 Digital freedom is not a default state

The easy assumption is that digital freedom is what you get when nobody interferes. Remove the rules and freedom appears.

That is not how it has worked. Digital freedom is a construction. It stands on three things, and it needs all three at once:

1. Sovereignty. Not being dependent, technologically or legally.
2. Reliable rules. Rules that are clear enough to plan against.
3. Real trust. Not declared trust. Demonstrated trust.

Take any one away and the other two stop working. That is the part most of the public debate misses, because it usually argues for one at the expense of another.

☁️ Sovereignty: the question behind the cloud question

Digital sovereignty is on the programme as a theme in its own right, which is telling. The question underneath it is simple: are we actually sovereign?

The cloud debate in Europe is usually framed as a question about where a data centre sits. That framing is comfortable and it is incomplete.

The real question is who can compel access, under which legal order, and whether you would find out. A data centre inside the EU, operated by an entity subject to a non-EU legal order, is a geography answer to a jurisdiction question.

And "digital assurances" from a provider are exactly that. Assurances. They are worth something. They are not the same as an architecture where the access does not exist to be compelled.

Three sovereignty questions worth more than any certification:

• Where do the credentials and the keys physically live, and who holds them?
• If we want to leave, what exactly do we take with us, and how long does it take?
• Can the provider change the terms of service under us, and what is our position if they do?

Sovereignty is not a flag on a map. It is revocability, portability and the absence of a single point of compulsion. 🔑

⚖️ Rules: regulation is not the opposite of innovation

Artificial intelligence sits on the programme alongside it, including the current state of play on the EU AI Act and the legal classification of AI agents specifically.

There is a reflex in parts of the technology industry that treats every regulation as a brake. Sometimes it is. But the reflex gets the causation backwards.

Rules that are clear create the conditions for investment. A regulated market is a market where a buyer can sign a three year contract without fearing the ground will move. An unregulated market is one where nobody at board level will approve the budget, because nobody can say what the exposure is.

What slows innovation is not regulation. It is regulatory uncertainty. Those are different problems with opposite solutions.

The honest version is this: bad rules do slow things down. Rules written without understanding the technology, rules that require evidence nobody can produce, rules that change every eighteen months. The answer to those is better rules, not fewer.

The agentic case makes this concrete. An AI agent that takes an action on a system of record is not a chatbot. It needs an identity, a bounded permission, an approval point and an audit trail. Every one of those is a legal requirement and an engineering requirement at the same time. That is not regulation fighting innovation. That is regulation describing the architecture. 🤖

🤝 Trust: the part that cannot be declared

The third element is the one organisations most often try to shortcut.

Trust is not a badge on a website. It is not an assurance in a contract. Trust is what accumulates when a system behaves predictably under conditions nobody designed for.

In practice it comes from unglamorous things:

• The system says what it does not know, instead of guessing confidently
• A human can see what happened and why
• When it fails, it fails loudly rather than quietly
• Someone is accountable, and that someone is named

Anonymity and freedom of expression are both on the programme, and both are trust questions wearing different clothes. How freely and how safely people move around the internet is decided daily, by design choices most users never see.

🔄 Why the three cannot be separated

This is the part worth sitting with.

Sovereignty without rules produces fragmentation. Every country building its own stack, nothing interoperating, and a smaller market for everyone.

Rules without sovereignty produce compliance theatre. You can write the best AI governance framework in Europe and still have no control over the infrastructure it runs on.

Either one without trust produces a system nobody uses. Adoption is where governance either proves itself or is revealed as paperwork.

Freedom and responsibility are not a trade-off. They are the same structure described from two sides. 🧩

🏗️ What this means for organisations actually building

A conference like this one is where the legal order for the next decade gets argued out. Most of what is decided there arrives in engineering backlogs eighteen months later.

So the practical reading, for anyone deploying AI inside a European organisation:

• Design for the jurisdiction question now. Not because a regulator will ask tomorrow, but because retrofitting a data perimeter is far more expensive than building one. Controls designed in early are cheaper to build and much easier to evidence than controls bolted on against a deadline.
• Treat the audit trail as a product feature. It is what lets you expand the system's authority later. Without it, every extension needs a new argument from zero.
• Put the human where the decision is reversible. Not everywhere, which stalls the system. At the points where being wrong is expensive.
• Keep the exit real. If you cannot describe how you would move off a platform in a page, you do not have an exit. You have a hope.

🎯 Where we stand

We build and operate agentic AI systems for European organisations, across German and Indian teams. The questions on this conference programme are the questions our clients raise in the first meeting, usually in a less precise form.

Our position has not changed and this week has reinforced it: model capability is no longer the binding constraint for most enterprise workloads. Governance, unit economics and ownership are.

An agent that can read is a demonstration. An agent that can write needs four things: an identity, a bounded permission, an approval point and an audit trail. Sovereignty, rules and trust are not abstractions in that sentence. They are what those four things are built out of.

🚧 What we will not claim

We are not lawyers and nothing here is legal advice. These are our notes on a public conference programme and our reading of what it means for people who build.

We also will not claim that governed systems are free. They are slower to start and they cost more at the beginning. What they buy is the ability to expand scope later without renegotiating trust from scratch, which is where ungoverned pilots usually die.

And the legal position on several of these questions, agent liability in particular, is genuinely unsettled. Anyone telling you otherwise is selling something.

📅 Talk to us

Book a 30-minute review. We will tell you which of your AI use cases can stand up to the jurisdiction question today, and which cannot.

Data privacy, GenAI and advisory: mejuvante.ai/enquiry-1