An AI-generated audio recording of the late celebrity chef Anthony Bourdain's voice - which appeared in a recent documentary about his life - has kick-started a public debate about posthumous privacy.
Do we deserve privacy, even after we die? Is our data still ours once we're gone, or an extension of our identity that should also be laid to rest? Can private data exchanges be made with spousal or family permission?
The audio recording
The recording that has led to several articles being published in the US press is a 45-second audio clip that appeared in Roadrunner, a Netflix documentary about Anthony Bourdain's life. The clip itself, based on many hours of Bourdain speaking, is a completely new piece of audio that was produced entirely using AI technology.
Quite controversially, at one point in the documentary, the AI-generated voice reads out a previously private email, one that was written before the chef's suicide in 2018.
Morgan Neville, the director, claimed in an interview with men's magazine GQ that he had checked with Bourdain's widow whether he could use AI to recreate her late husband's voice. But she has said since that she never signed off on the decision and never gave her permission on Anthony's behalf.
Why is it so controversial?
This isn't the first case of someone being brought back, in some way, shape, or form, from beyond the grave using artificial intelligence. 2D and 3D renderings of dead celebrities - the most famous, recent example being that of Kim Kardashian's deceased father - have been around for several years now and are becoming increasingly sophisticated and lifelike as time goes by.
There are three ethical questions that preside over this area. The first is the broadest: is it ever right to publish text or audio someone wrote or recorded after they die if you never obtained permission before they passed? Such a task could be completed without AI, of course. Some believe the permission of close family and friends - which the director of Roadrunner claimed he received - is all that is needed to greenlight a posthumous invasion of privacy; others believe consent can never be granted, even by a trusted third party.
The second issue is whether it is ever permissible to use artificial intelligence to present a dead person's data or information in a way they never intended. This transfer to a 'new medium' happened in the Bourdain case. The chef once wrote the email that forms the basis of the 45-second clip featured in the documentary, but what if he never intended it to be read aloud, or with the inflection and cadence generated by the AI in use.
The third concerns the production of entirely new material produced using things like Deepfake technology, and whether it's ever morally permissible to create, for instance, an entirely new audio recording that has no basis in something the deceased person in question ever said.
Posthumous privacy and post-death data rights
The issue of posthumous privacy, and whether individuals have a right to it as such, invokes a number of philosophical, ethical, and legal concepts and is a rather new frontier in terms of rights. Prior to the last decade or two, it was only celebrities that had their existences documented in a detailed enough fashion for this conversation to take place.
Nowadays, whether we like it or not, most people now have an incredibly deep and complex digital footprint that will outlive them. There will be data about our lives that persists online for years and years after we die.
One problem is that it's unclear who owns this data. There's a reason we write wills before we pass, and divvy up all of our worldly possessions; it feels intuitively wrong to suggest someone dead can claim sensible ownership over a material object, but it's also odd to suggest they don't own it either. Making these decisions before we die is a way to avoid arguments over what a deceased person would have wanted. But the fate of the deceased's social media accounts and old emails are rarely afforded the same level of pre-planning.
To wade deeper into the ethical quagmire, it also makes sense to ask what role the companies that hold and host our data and information play when it comes to ownership - especially public data like statuses or tweets. Sites like Twitter expressly state in their privacy policy that users own their tweets. However, the debate about ownership has largely centered around online abuse, which social media companies are all too keen to shirk responsibility for - but this may change if, for instance, if it can be shown to be financially lucrative.
Another quandary is what we're actually doing when we generate AI versions, or new iterations, of a person, saying things they never said when they were alive. Are we creating a post-death extension of a person, or is an AI-generated 'version' of a dead individual in fact a distinct entity in itself, only inspired by, or made in the image of, the person that was once alive?
A civilizational question
Right now, there are no legal protections to stop you from being digitally resurrected, no way to 'opt-out,' and no real societal precedent for securing your posthumous privacy.
But there is a rapidly growing graveyard of deceased individuals, outlasted by their digital footprints. Facebook may have almost 5 billion 'deceased' accounts by the end of the century, and this bank of data will take on an increasingly important role for historians, sociologists, and anthropologists.
What happens to all of our data after we die, then, is a question we may have to answer sooner rather than later.