Machine Made Me

A Deepfake bill in Denmark, an AI-generated victim addressing an Arizona court, and a digital artist receiving official credentials in Beijing are all testing the same question: when a synthetic version of a person begins to speak and act, what relationship does it bear to the self?

Three stories have appeared in quick succession.

In Denmark, the government has proposed legislation to stop realistic digital copies of a person’s appearance or voice being made and shared without consent.

In Arizona, an appeals court ordered a man to be resentenced after an AI-generated version of the person he had killed appeared at the original hearing and addressed him from beyond the grave.

In Beijing, an AI-powered virtual idol called Yuri has received a Digital Artist Card, giving her operators a regulated route into commercial performance.

These look like three different stories: one about deepfakes, one about grief and justice, and one about virtual influencers.

They are really one story.

The machine-made self is entering the world.

The right to control your double

Denmark’s proposed law would prohibit the digital copying and sharing of lifelike representations of a person, including their appearance and voice, without consent. It would also protect artists from unauthorised imitations of their performances, while preserving exceptions for parody, satire and social criticism.

It has often been described as giving people copyright over their own face, body and voice. Strictly speaking, the more important development is not that our features become property. It is that the law begins to recognise an enforceable relationship between a person and their synthetic likeness.

Until now, a copied face or cloned voice has often fallen awkwardly between existing categories. Is it an image, a performance, personal data, an impersonation, a fraud or an assault on identity? Copyright law protects works. Privacy law protects information. Defamation law protects reputation. So, none quite captures the peculiar violation of encountering a version of yourself doing something you never did.

The Danish proposal starts from a different premise: a machine-made likeness is not simply content. It remains connected to the person from whom it was derived.

That connection is the contested territory. If the replica looks like me and sounds like me, but acts without me, what exactly has been taken?

When the dead speak in court

The Arizona case makes the question more difficult because the original person could no longer consent or object.

Christopher Pelkey was killed in a road-rage shooting in 2021. At the sentencing of Gabriel Horcasitas, Pelkey’s family presented an AI-generated video in which a digital version of Pelkey addressed the man convicted of killing him and spoke of forgiveness. His sister had written the words. The avatar delivered them with Pelkey’s appearance and voice.

The trial judge responded warmly to the video. But in September 2026, the Arizona Court of Appeals ordered a new sentencing hearing. The conviction remained in place, but the court found that reliance on the synthetic statement had made the sentencing process fundamentally unfair.

The problem was not merely that the video had been generated by AI. It was that a machine-made Pelkey was presented as the “true representation” of who he had been, while speaking words he had never said.

His family’s intentions were compassionate. The video was part of their grieving and an attempt to make the victim present in a process dominated by the offender. Who has the right to write the dead? Who may decide what they would forgive, condemn or request? At what point does remembrance become authorship?

A photograph preserves an appearance. And a recording preserves something once said. But a generative replica can produce new speech and new actions indefinitely. It does not merely keep a memory. It manufactures a future performance. Exactly what I rasied in my book in 2021 in the final chapter ‘Destroying You.’

A synthetic identity enters the system

The third story moves in the opposite direction. Instead of a human being reproduced as a synthetic version, a synthetic character is being given the credentials needed to operate in human systems.

In June, the Beijing Economic-Technological Development Area issued Yuri, an AI-powered virtual idol, with China’s first Digital Artist Card. The card is part of a regulatory sandbox for commercial performances by digital humans. It provides a compliance route through which intellectual-property rights, content safety and operational responsibility can be checked and monitored.

Some reports have described this as China issuing a government identity card to an AI person. In fact, Yuri has not been granted human legal personhood. She is operated by a team, and the credential exists partly to make that operation traceable and governable.

But the direction of travel matters. A synthetic entity has been made sufficiently legible to the system that it can be certified, monitored and admitted into commercial life. It can maintain a persistent name and appearance, attract an audience and enter relationships with businesses through its operators.

The representation has become operational.

I asked these questions in 2021

None of these questions arrived with generative AI.

In my 2021 book, The Future of You, I explored what happens when identity can be captured, inferred, copied, recombined and performed by machines. I asked who would own and control our digital likenesses, what authority our replicas might acquire, and whether the versions of us produced by technology would remain representations or begin to function as selves in their own right.

That July, in Virtually Immortal* for BBC Radio 4’s *Four Thought, I turned specifically to resurrection AI. I asked what happens when the memories and data of the dead are used to create virtual assistants that continue to speak. Who, or what, is being memorialised? Is it the person, or something else assembled in their image?

At the time, these questions were often treated as speculative. Five years later, lawmakers, judges and regulators are being forced to answer them.

The progression is becoming visible:

representation → reproduction → participation

The digital likeness begins as an image of someone. It becomes a generative version capable of producing new words and actions. Then it enters a courtroom, a contract, a commercial partnership or an intimate relationship and begins to have effects in the world.

Once that happens, authenticity is no longer the only issue. A synthetic self does not have to be mistaken for the original in order to exercise influence. Everyone in the Arizona courtroom knew that the video was generated. It still affected the proceedings. Yuri’s audience knows she is virtual. She can still attract loyalty and economic value. A labelled replica can still act upon us.

This is Me:chine

I call this condition Me:chine.

Me:chine is not “man versus machine”. It describes the meeting of the machinable self, the part that can be recorded, modelled, predicted and reproduced, and the unmachinable self, the part that cannot be fully captured by computation.

The colon matters. It holds open the space between the person and the machine-made version. That space is not fixed. It is being negotiated by families, companies, platforms, courts and governments.

These three stories reveal four tests that will increasingly govern it.

Consent. Who authorised the replica to exist and to act?

Authority. What may it legitimately say or do in the person’s name?

Continuity. In what sense, if any, is it connected to the original self across time or after death?

Accountability. When it causes harm or acquires influence, who is responsible: the model, its maker, its operator, its owner or the person it resembles?

The Danish proposal concentrates on consent. The Arizona judgment exposes the problem of authority. The Beijing pilot addresses accountability by making the operators of a synthetic character visible to the system. Continuity remains the most philosophically difficult because no credential can prove that a generated version is the person it continues to perform.

This is why the argument cannot be reduced to detecting fakes. Detection helps us establish how something was made. It cannot tell us whether a synthetic performance was authorised, whether it should be permitted, or what relationship it bears to the person whose identity it uses.

Nor is this only about protection from bad actors. Some of the hardest cases will arise from affection, creativity and care. A grieving family may want to hear a loved one again. An artist may license a younger digital double. A person may deliberately train an agent to speak and act for them. The copy may be wanted, useful and intimate.

The question is not whether machines can reproduce us. They can.

The question is who gets to create those versions, and on whose terms.

When a machine-made version of me speaks, transacts or participates in the world, is it my property, my proxy, my continuation, someone else’s interpretation of me, or a new synthetic participant altogether?

That is no longer a question for a distant technological future. It is becoming a legal and lived condition now.

The machine-made self has entered the world. We are only beginning to decide the terms on which it may remain.