Reflections on Digital Identity in the age of AI

In the ever-evolving era of artificial intelligence, the sixth season of Netflix’s Black Mirror casts a timely spotlight on a future where digital identity is exploited to the extreme. Joan is Awful, the new season's first episode, explores what happens in a fictional future without appropriate safeguards and standards and the consequences for individuals who lose control of their digital image rights, robbing them of their identity. Just last week, the Sag-Aftra union joined the Writer’s Guild of America by calling a strike to re-negotiate their terms with studios, with the industry’s adoption of AI and digital representation a major point of contention. 

As AI continues to populate our professional and personal lives, protecting your likeness and data has become an increasingly real concern. Generative AI tools such as Midjourney, Runway, Eleven Labs, and many others have already demonstrated how easy it is to create realistic images of celebrities and likenesses of their voices, blurring the line between genuine and artificial. In some instances, they have managed to deceive individuals into believing they’re genuine (see the Balenciaga Pope and the early Trump arrest shots). As we quickly move from static AI images to AI video, the potential to misuse and mislead people only increases.

The promise and pitfalls of digital doubles and deepfakes

In the realm of filmmaking, visual effects teams have long created digital doubles with vast budgets and mixed results but two recent films have harnessed the power of AI to rejuvenate their lead actors with mind-blowing authenticity. Indiana Jones: Dial of Destiny and Robert Zemeckis' Here, featuring Harrison Ford and Tom Hanks respectively, have leveraged AI technology to seamlessly de-age or age their stars. While it’s still Ford’s performance, AI tools are used to recreate his youthful self using existing footage. Here achieves the same photorealistic results for Tom Hanks, using a generative AI tool called Metaphysic Live, enabling real-time, high-resolution face swaps that grant the director instant feedback during filming.

Metaphysic, the team behind the technology and the viral @deeptomcruise account, are leading the charge in bringing this technology to the film industry. When asked about the complications surrounding the use of AI-generated likeness, Metaphysic co-founder Martin Adams states “We insist upon full and informed consent from them [the film studios] contractually, but also from any talent that they’re looking to have the technology applied to.” (TVB Europe).

Film is far from the only industry to be pushing the envelope with state-of-the-art generative AI. Metaphysic stunned audiences last year when they brought Elvis and the cast of judges on America’s Got Talent to life for a televised performance. In our piece Redefining Live: Mixed Reality & the Music Industry, we explored the move towards these virtual concerts that both revive artists of the past and provide new tools for living artists to engage audiences, like the R.A.V.E. (Realtime Audio Variational autoEncoder) model used by vocalist Jen Wang last month (New York Times).

As a prime example of how this technology can harm artists rather than help them, the AI-generated ‘Heart on my Sleeve’ featuring replicated versions of Drake and The Weeknd shook up the industry last April, receiving millions of plays on TikTok, Spotify, YouTube before being taken down. While most oppose this infringement of digital likeness, Canadian artist Grimes embraced the move with the launch of elf.tech, creating an open-source of her AI-generated voice with the clause of a 50/50 split on track royalties. What’s seen as a novelty by most has also drawn parallels to the effect that sampling and file-sharing have had on music production and consumption in the past.

In politics, social platforms have been bombarded with AI-generated clips of speakers (most notably US President Joe Biden) discussing everything from hip-hop and video games to spouting harmful anti-trans rhetoric (later debunked by PolitiFact). While most can identify the content for what it is, the speed at which the videos can be created presents a massive threat to news cycles and existing methods of authentication (see Eleven Labs’ demonstration of AI voice technology with Leonardo DiCaprio).

Not dissimilar from the doctored videos of politicians and podcasters, celebrities have begun fighting back against the misuse of their likeness in AI-generated adult content. Apps like DeepNude allow users to create non-consensual content powered by AI technology. Although removed from the app store one day after going live, it further demonstrates the potential harmful effect of the ungoverned, unprotected use of AI tools and digital likeness (Vox).

Harnessing your digital identity - and mitigating risk

Last week we saw the start of the Sag-Aftra strike, involving 160,000 actors in the US. Part of negotiations is the issue of ownership of an actor’s likeness if it is reproduced by AI and where the use of actors is threatened by AI in the future.

The only way to address these concerns is to comprehensively discuss the role that governance, data protection, digital image rights, authentication, etc. play in protecting people’s digital identities. It’s imperative that people have a clear understanding of and a way to distinguish between what is real and what is artificially generated.

At Dimension, we collaborate with celebrities and artists to create digital avatars with photoreal likeness. These virtual humans are the culmination of state-of-the-art production techniques across 3D scanning, real-time rendering, and AI. Eva Herzigova’s MetaHuman, created in collaboration with talent agency Unsigned Group, and Eva herself, is a prime example of a celebrity being on the front foot and seeking to be in control of their future digital identity. Eva’s digital avatar becomes an extension of her brand that opens up new opportunities for her to creatively engage with brand partnerships and work without limits.

These are uncharted waters, and we’re acutely aware of the profound ethical responsibility that comes with creating believable virtual humans. To this end, we foster collaboration with talent agencies, celebrities, and rights-holders to best develop and deploy their digital identity, facilitating control and ownership of their intellectual property.  

We also explore how AI itself can be employed as a guardian to ensure the accurate representation and safeguarding of authentically crafted content, mitigating the risk of misuse. Intermediary platforms like AI Incident Database and Deepware offer a glimpse into how we’ll be able to track and catalogue this misuse.

As these novel technologies advance and the boundaries between real and virtual continue to blur, Joan is Awful serves as a cautionary tale.