In Case of Voice

I started by looking at how the relationship between a person and a recorded voice has changed.

For a long time, recording meant preserving something that had already happened. Over time, the thing being preserved became larger: recordings, photographs, messages, posts, interviews, enormous archives of ourselves.

Then another possibility emerged: using those materials to construct something that could speak back.

Underneath the fictional Voice Clause is a real, unresolved policy landscape.

In 2024, the U.S. Copyright Office published the first part of its report on copyright and artificial intelligence, focused on digital replicas. It identified gaps in existing U.S. protections around unauthorized digital replicas of people’s voices and likenesses, and considered whether new federal legislation was needed. (U.S. Copyright Office ↗)

This case study is not a legal analysis, and the project is not legal advice. A few careful qualifications matter:

• The project does not claim that a person “owns their voice” in a simple, universal sense.
• Existing protections vary by jurisdiction and may involve different legal interests in recordings, likeness, publicity, contracts, estates, and digital replicas.
• The fictional Voice Clause is a speculative design proposal, not a legal instrument.

The point is simpler: the questions the prototype dramatizes are already being argued about in the present.

The course distinguishes a forecast from a scenario. A forecast is a plausible direction built from present signals. A scenario is a specific situation inside that broader future — it puts you in a world and asks what it would feel like.

I wasn’t trying to predict what voice technology will look like in 2047 or 2081. I wanted a working hypothesis to test:

The first version of the idea was simply “write a voice clause.”

The scenario work made me more interested in what would happen after the clause was written. So I kept the document fixed and changed the circumstances around it: four requests, arriving between 2047 and 2081, each one harder to answer than the last.

At the center of the prototype is a fictional Voice Clause, written in 2031. It covers whether the voice can be recreated, who can authorize it, what it can say, where it can be used, commercial use, duration, and what the voice must never say.

Rather than reproduce the whole artifact here, two excerpts matter most — one a rule, one an admission that rules run out.

In 2047, the daughter’s request fits the permitted setting; the ambiguity is in the message itself. In 2063, the system is answering a question the person never answered — an echo of Jamais Cascio’s futurist work in my course materials, which imagines simulations of the dead eventually extrapolating responses to new problems. In 2067, the use is allowed but the categories no longer fit the world. By 2081, the clause has expired. In this scenario, authority passes to the estate after the clause expires. But the context needed to interpret the person’s intent is gone.

The document is still legible. Its author isn’t available to explain it.

In the first version, I asked the visitor to decide whether each request was ALLOWED, PROHIBITED, or AMBIGUOUS.

It made sense as an interaction. Then I realized it was undermining the premise: the visitor inhabits the person who wrote the clause, and by the time the requests arrive, that person is no longer alive. They don’t get to adjudicate what the document means.

The revised experience then shows the document assessment — which provisions are clear, which are prohibited, and which require interpretation. The people and institutions involved can disagree. The visitor is there to see the dispute, not resolve it.

That small change made the whole project more coherent.

The artifact uses Bureaucratic Brutalism because the visual language is part of the idea. The interface looks like an archival record, an estate system, a legal document, a government form. There are no futuristic graphics, avatars, or voice effects.

I also deliberately kept the prototype silent. I didn’t want to build a demonstration of voice cloning; I wanted the visitor to think about what a voice means when the technology itself becomes ordinary.

There was another question I hadn’t expected to encounter through the project. I started thinking about my own voice.

Portuguese is my first language, and I speak English with an accent. That made the word voice feel less straightforward.

If my voice were reconstructed decades from now, what exactly would need to survive for it to still feel like mine?

I started with a question about consent.

The project became more interesting when I started thinking about interpretation.

A person can write down what they want. They can name who should decide. They can try to anticipate future uses.

The difficulty is that someone else may eventually have to interpret those instructions in circumstances the original person could never have imagined.