AI Copycats: Actors Embrace the Challenge

Voice actors from Latin America to Africa are working together to get better contracts and laws that keep their voices safe from AI

Glenn Burgess

Armando Plata, a voice actor, is surprised to hear his voice being used for various purposes, like promoting a shopping mall in Bogota, narrating an adult movie, or advertising big bank, without his consent or compensation.

Armando Plata realised his voice had been copied by artificial intelligence when he noticed a mild, robotic tone in the cloned recordings. This took away his artistic choice and vocal rights, which are vital to him as a veteran actor with a 50-year audio career.

Plata, who owns a deep and melodious voice and serves as the president of the Colombian Association of Voice Actors, believes that his voice is one of the most frequently cloned and used artificially in Spanish.

In response, Plata is working with voice actors from across Latin America to create legislation that protects the “right to own one’s voice.” This effort is part of a global movement for human rights in the face of the rapid growth of AI technology.

Artists worldwide, from South Africa to Europe, Japan to the United States, are uniting to safeguard their livelihoods and individuality against the impact of AI technology that can replicate their voices.

The story began innocently enough. Two decades ago, Plata participated in a paid project, recording text-to-speech for a company. Unbeknownst to him, this company later sold his recorded voice to an AI software company.

While his voice became popular, it didn’t bring him any financial gain because, as is customary in the voiceover industry, Plata had not signed a contract. This meant he had no legal recourse for pursuing lawsuits.

Plata believes that in the future, they may have the ability to sue companies and advocate for class action lawsuits. However, the first crucial step is to have governments recognise the concept of owning one’s own voice.

Human right to a voice

In 2023, around 500,000 video and voice deepfakes are expected to be shared on social media platforms, as reported by DeepMedia, a company specialising in synthetic media detection.

The cost of cloning a voice has dropped significantly. It used to require an investment of $10,000 for server space and AI training. Nowadays, startups offer this service for just a few dollars, according to DeepMedia.

Notable cases of voice appropriation include the use of Morgan Freeman’s voice and likeness in a fake video to criticise President Joe Biden in April 2023. The affordability of AI clones, which can cost about half as much as hiring a human voice artist, makes this technology enticing.

For instance, one voice actor listed on the Voices.com marketplace in Colorado charges $500 for a 60-second radio ad, while the AI equivalent costs $200 per minute, approximately $1 per word, with some options being even more cost-effective.

Speech-generating AI, like Microsoft’s VALL-E language model, functions by analysing large volumes of data to understand how people speak. It then uses a neural network algorithm to mimic human speech patterns and characteristics.

In Chile, after the emergence of AI-powered audio production companies this year, the national association of voice actors engaged with lawmakers to discuss the concept of voice ownership as a fundamental human right.

Similarly, voice actors in Colombia have initiated a legislative project to establish the human voice as personal heritage.

These legislative efforts aim to provide a foundation for future regulations, potentially requiring the use of audio watermarks in all content generated using synthetic voices.

Unlike copyrighted works that are preserved in a tangible form, such as paintings or digital files, voices currently do not have legal protections.

While some countries have rules against deepfake videos featuring celebrities, there are no specific laws governing vocal deepfakes.

Personal information?

In Africa, voice artists are working to protect their interests, although the diverse array of regional accents and languages in the region presents challenges for AI voice models, according to Andrew Sutherland, a South African sound engineer and voice artist.

One potential avenue for protection is the South African Protection of Personal Information Act, which prohibits the collection, processing, or storage of personal data, including voice, without consent.

Sutherland suggests that a voice could be considered personal and sensitive data, as it can reveal information like a person’s class or age, and could be protected on those grounds.

The South African Guild of Actors is actively advocating for government policies that safeguard performer rights, a strategy also being pursued by Arts Workers Japan, the main industry body for freelance performers in Japan.

While Tokyo’s copyright laws currently favor AI, the city may need to introduce protections for aspiring actors, as AI can instantly generate similar content, potentially hindering their future success, according to Michihiro Nishi, a partner at law firm Clifford Chance.

Megumi Morisaki, an actor and president of Arts Workers Japan, notes that Japan has sought new legislation within the G7 while relying on outdated laws to address emerging challenges. This has resulted in limited protection for artists in the digital age.

Who rules in AI world without borders?

Artists worldwide are looking to the European Union’s AI Act, which categorises AI tools by potential risk, to establish a global foundation for the use of synthetic voices.

Tim Friedlander, President of the National Association of Voice Actors (NAVA), highlights the global nature of the industry and the challenge of determining voice ownership when collaborators span different countries. He also notes that AI legislation in the United States seems to be at a standstill.

NAVA is advocating for AI voices to be classified alongside deepfake images in a high-risk category. This is crucial because it impacts not only well-known voice actors but anyone who has recorded audio, potentially exposing them to scams or blackmail.

Spanish-speaking voice actors have formed the United Voices Organisation, focusing on fair compensation and ethical use of voices in AI-related projects.

While organisations like NAVA and the Screen Actors Guild-American Federation of Television and Radio Artists (SAG-AFTRA) have contractual safeguards against AI abuse, less than one-fifth of industry workers are unionised. In South Africa, voice artists are considered freelancers and lack union representation for rights and fair compensation standards.

Urvashi Aneja, the founding director of Digital Futures Lab, emphasises the importance of protections in informal markets, particularly in Global South countries, where informal and precarious work is increasingly common.

Until global legislation is in place, advocates hope that platforms will pay recognised voices rather than startups using AI copies generated from web-scraped data.

The ethical aspect of AI is gaining prominence, and organisations are now seeking ethical sources for voice acting, even if it comes at a higher cost than AI alternatives.

Armando Plata, a voice actor whose voice was cloned without consent, believes that AI can become a new source of artistic income if properly regulated. He envisions a future where his family can continue to benefit from his voice even after he is gone, allowing his legacy to endure.