“Again!” the producer exclaims in the dimly lit voice recording studio, his frustration evident as he scowls at the monitor displaying a waveform of the latest recording.

In the past, such a moment of discontent would have meant asking the voice actor to stay late or juggle schedules for a re-recording, a painstaking process marked by the click of the studio’s clock and the slow churn of retakes.

However, in 2024, such hassles are relics of the past; the studio is now home to a suite of advanced AI TTS (text-to-speech) and voice cloning tools that have introduced a new level of convenience.

Nowadays, with just a 5-minute sample of someone’s voice, it can be stored digitally and used however one wishes – whether it is resurrecting a deceased person’s voice or restoring someone’s lost ability to speak. However, as this groundbreaking technology becomes more embedded in our lives, it brings complex intellectual property issues for the creatives.

“You can see how it can be extremely beneficial, but with the power comes the ability to cause harm,” said Ted Chwu, a partner at Bird & Bird in Hong Kong.

“How do you regulate it? How do you police it?”

Ted explained that the legal field is beginning to grapple with these issues. Nevertheless, even at this early stage, significant attention is already being given to the adverse effects that have been uncovered.

The rising concerns over AI and intellectual property

This technological leap has not come without its challenges. While many are concerned about jobs being replaced by AI technology, the current debate is not just about the clash between humans and machines. Still, it focuses on how tech companies and regulations are handling the rights of human workers amidst technology that is advancing at a pace often outstripping the development of regulatory measures, as well as new questions about whether the laws governing creative expression, such as copyright, are a source of protection.

AI technology made headlines again on July 26, 2024, when hundreds of Hollywood video game performers went on strike in Los Angeles, protesting the lack of worker protections against the unregulated use of artificial intelligence.

The strike comes after nearly two years of negotiations with gaming giants, including Activision, Warner Bros, and Walt Disney Co., over a new interactive media agreement. SAG-AFTRA (Screen Actors Guild-American Federation of Television and Radio Artists) negotiators say gains have been made over wages and job safety in the video game contract.

However, the two sides remained split over the regulation of generative AI. According to a SAG-AFTRA press release, the most significant sticking point in negotiations is the unrestricted use of artificial intelligence. Video game performers want protection over the use of their likenesses and voices.

“We’re not going to consent to a contract that allows companies to abuse AI to the detriment of our members. Enough is enough. When these companies get serious about offering an agreement our members can live – and work – with, we will be here, ready to negotiate,” SAG-AFTRA president Fran Drescher said in a statement.

These performers are not the first to raise alarms about undermined rights when AI is brought into the conversation. Artists have long fought to protect their rights, and the advent of fast-developing AI has intensified this struggle.

This issue is exemplified by Scarlett Johansson, who sued OpenAI and claimed that the company’s AI voice assistant closely resembles her voice after rejecting an offer from Sam Altman, CEO of OpenAI, who wanted to hire her to voice ChatGPT. She argues that using this AI voice infringes on her rights and seeks to hold OpenAI accountable.

The technology’s ability to blur the line between reality and fabrication makes it especially dangerous, as people may struggle to distinguish between the two as AI advances. The growing concern is that lack of regulation on convenient access to technology may lead to duplicates or loss of identity.

“For instance, the digital appropriation of personalities and voices means the individuals are at greater risk in losing control over their digital identity and how their likeness is used,” said Stanley Lai, head of the intellectual property practice and co-head of the cybersecurity and data protection practice at Allen & Gledhill in Singapore.
 

“This can, in turn, cause emotional and psychological harm where there is malicious misuse, causing reputations to be ruined and individuals to be harassed,” he added.

Depending on the jurisdiction, TTS AI technology can raise concerns such as civil rights issues, deceptive practices, false advertising, unfair competition, and unjust enrichment.
 

For instance, “such systems may provide output based on limited data obtained from a peculiar region of a country or some wrong imitations of accent portrayed in a movie which may not be providing a correct and complete picture of how people from a country speak,” said Pravin Anand, managing partner of Anand and Anand in Noida.

Who owns your voice?

This question raises intriguing issues: the notion that a voice can be owned and that someone other than the speaker might claim ownership of it can seem perplexing.

Complexities directly impact the voice acting and performing industry, as artists rely on vocal abilities to create distinct characters. Their craft involves maintaining trained technique and vocal health for compelling performances. When their voices are used publicly without consent, it undermines the value of their work and erodes their control over creative output.

“A major legal issue surrounding TTS is the potential infringement of a person’s right to control the unauthorized commercial use of their identity, such as name, image or voice. Using sound recordings featuring a person’s voice to train a TTS model may violate the copyright in the recording. Data protection issues are also relevant, as a person’s voice may amount to personal data. In some jurisdictions, this is considered biometric data,” said Amita Haylock, a technology, media, and telecom partner at Mayer Brown in Hong Kong.
 

Haylock explained: “Many artists and voice actors may not have a good grasp of intellectual property laws and therefore may not appreciate the rights they are signing off on when they agree to work with TTS technology developers or providers. The potential risks are that artists and voice actors could lose control of how TTS companies use their voices, what content their synthesized voice appear in, and not receive fair compensation.”

Remie Michelle Clarke, Microsoft Bing’s voice

A phone call rings for Remie Michelle Clarke, an Irish voice actor known for her work with Microsoft, Audible, Mazda and Mastercard. She is currently the voice of Microsoft’s search engine, Bing.

“I wanted to know how it worked for you, creating an AI voice, was it worth your while?” Clarke recalled the conversation with a sound engineer she worked with for two decades.

Then Clarke learned about Revoicer, a company specializing in AI-driven synthetic voice technology. She was shocked and confused, having never heard of Revoicer, and has found herself at the centre of a contentious issue involving unauthorized voice cloning.

There it was; her voice was listed on Revoicer’s website under the alias “Olivia,” paired with a photograph of a grey-haired woman who bore no resemblance to Clarke. Despite a somewhat altered voice quality, she knew it was unmistakably hers.

Clarke contacted a media law expert and was informed that her voice was not a copyrightable asset under European law. Instead, it is seen as the interpretation or performance of copyrighted text, complicating the legal pursuit against unauthorized use.

As Clarke sought clarification, she attempted to contact Revoicer but received no response, and the company lacked a physical address. She continued to trace the source of her voice to determine where the breach occurred. Clarke’s contract with Microsoft years back is likely to have granted third parties the right to use her recordings indefinitely, a clause that her legal counsel admitted made her options for challenging the misuse of her voice relatively limited. Despite her efforts, Clarke has received no direct communication from Revoicer or Microsoft, even after numerous media inquiries.

“I’ve had no correspondence directly from any of the parties involved, despite extensive media attention,” Clarke tells Asia IP. Her frustration is evident as she navigates the murky waters of voice cloning and its implications for her career and the industry.

During the Covid-19 pandemic, remote work led to a surge in home studio upgrades among voice actors, with many transitioning to remote setups. A 2021 Voices.com survey shows that around 70 percent of voice actors upgraded their home studios during the pandemic. The influx of non-professionals has driven down rates, creating additional competition and diminishing opportunities for seasoned voice actors.

“The work isn’t as good; the client often doesn’t want to pay as much. They know they can get you cheaper as a robot voice or simply someone cheaper,” Clarke tells Asia IP. This sentiment echoes concerns across the creative industry, where many artists face similar issues. The ease with which AI can clone voices and replicate artistic styles has raised significant questions about ownership and rights transfer.

Clarke parallels visual arts, where AI tools have begun replicating artistic styles and raising concerns about the unauthorized use of artwork in recent years. “In all other aspects of acting and the art world, there are protections for artists when it comes to reuse and resale,” she pointed out.

“I could do a job for the BBC in 2012, and, if in 2016, they wanted to sell it to Netflix, say, they would pay me a residual. It should be the same for generative AI. I don’t understand why they feel exempt from these protections,” Clarke said.

She argues that the unchecked growth of AI in creative fields threatens to undermine the value of artistic work and exploit artists. “They’re destroying creative industries in their drive for what they would call progress,” Clarke said.

“I wouldn’t call it progress, but I certainly call it profit for them.” Clarke’s situation highlights a broader issue in the creative industry: while high-profile artists may have the means to seek legal recourse, those without significant resources often find themselves with limited options for protection and compensation.

Many voice actors wonder whether they can hold their importance and value in the creative industry, competing against their digital doppelgänger and lower-cost voice actors. “You’re now competing against your robot voice for jobs; it has impacted the whole industry, the standard that we all try to uphold by holding our line when it comes to rates,” said Clarke. As she is taking action, she has contacted four lawyers who have given her other answers for the past year and a half.

Voice ownership beyond borders

As technology crosses borders, local laws often lag, creating confusion over voice ownership. In some areas, personality rights may protect voice but not copyright. The legal landscape varies widely, with different regions offering diverse levels of protection through personality rights, publicity rights, and copyrights. Legal actions depend heavily on the jurisdiction and specific case.

On April 23, 2024, the Beijing Internet Court ruled on the first case of AI voice infringement. The court decided against a defendant who used a voice actor’s voice to train generative AI without permission. As of August, neither party has appealed, and the ruling is effective. The new Civil Code of the People’s Republic of China includes a section on personality rights, explicitly protecting voices for the first time.

Senior Judge Zhao Ruigang stated: “It is important to note that, as a type of personality right, a voice has a personal and exclusive nature. Therefore, any individual’s voice should be protected by law. Authorization for the use of recorded products does not imply authorization for AI processing of the voice. Unauthorized use or permission granted to others to use the voice from recorded products without the rights holder’s consent constitutes an infringement.”

The court found that the defendant’s use of the plaintiff’s voice for AI TTS development violated the plaintiff’s rights, ordering the companies to pay Rmb250,000(US$34,500) in compensation. The AI-generated voice closely mimics the plaintiff’s vocal characteristics, making it recognizable.

Although the voice-acting company owned the copyright of the recording, this did not include the personality rights of the voice actor. The company did not have the right to license these rights to the other defendants.

The Beijing Internet Court case does not reflect the situation in all jurisdictions. As AI technology advances, the lack of specific legal frameworks means that cases often refer to existing laws and regulations, such as copyright infringement, performers’ rights, publicity rights, and data collection or privacy rights.

“To our understanding in China, we have some precedents mainly focusing on the outcomes of AI, but not training materials for AI,” said Frank Liu, partner and head of Shanghai Pacific Legal. This reflects the overall trend observed, primarily as expressed through judicial decisions.

“On one hand they want to protect copyright, on the other hand they cannot overly protect it, otherwise it may influence the development of AI; there should be a balance in between,” Liu added.

Imbalance in contractual agreements

As the demand for AI technology rises, voice actors struggle with contracts that often favour companies. These contracts may include vague “permitted use” clauses, giving companies too much control over voice recordings. They might also allow recordings to be transferred to third parties. Many actors also face inadequate compensation for the extensive use of their work. This problem is worsened by the power imbalance between companies and actors, especially for newcomers, making it hard to negotiate fair contract terms.

Justin Davidson, a partner at Norton Rose Fulbright in Hong Kong, advises that “TTS developers should obtain explicit consent to use a person’s voice for commercial purposes.” He explains that such consent should cover voice recordings’ personality rights and copyright issues. In jurisdictions where personality rights are not formally recognized, individuals may need to rely on alternative legal frameworks, such as copyright or trademark infringement, passing off, personal data protection laws, or defamation.

Adding to this complexity, Ian Liu, a partner at Deacons in Hong Kong, highlights the risks in regions without specific personality or portrait rights doctrines.

“In jurisdictions where there are no specific doctrines of personality or portrait rights, it results in risks that a voice actor may not be able to exert rights to their voice in absence of contractual agreements against another party’s use of their voice.”

Stanley Lai highlights the importance of fair contracts for voice actors. He emphasizes that actors should be cautious about overly broad “permitted use” clauses and ensure these are clearly defined and fair. Contracts should also restrict the transfer of voice recordings to third parties.

Lai stresses that adequate compensation and royalty mechanisms are crucial. However, he acknowledges that “voice actors often do not have the same level of bargaining power which the companies may possess.”

This imbalance is further underscored by the challenges newcomers face in the industry.

As Benjamin Choi, a lawyer at Oldham, Li & Nie in Hong Kong, explains, “I read a lot of things about bigger tech companies. They have resources to give incentives or they have resources to get licenses, and for a newcomer, you have much less bargaining power.”

The issue becomes more pronounced as TTS models advance and improve in mimicking original voices. The situation is further complicated when voice actors do not hold the copyright to their recordings, which employers or commissioning parties often retain. Liu explains that while the common law tort of passing off might offer some recourse, proving goodwill, misrepresentation, and resulting damage can be difficult.

“Famous voice actors may argue that a TTS service provider has passed off their goodwill given their reputation. For voice actors that have not established their goodwill and reputation, they may have a higher hurdle to rely on passing-off,” Liu notes. He underscores the difficulties faced by lesser-known actors in pursuing such claims.

Certain jurisdictions protect consumers against unfair contract terms from a business.

For instance, in Hong Kong, “a court may refuse to enforce unconscionable terms in a contract for sales of goods or supply of services where one party deals as a consumer. However, the question of whether the term is unconscionable is very often not clear-cut,” said Davidson.

He explains that if an AI company’s terms permit a user’s voice to be used to improve the existing service and to develop new services, the voice cannot be used on a standalone basis for commercialization.

“On the other hand, if an artist or a voice actor provides his/her voice recording to a TTS developer in a capacity other than as a consumer (e.g., if the TTS developer invites the artist or voice actor to provide voice recordings specifically for development of a TTS AI system), then it is possible that the TTS developer may want to acquire the right to use an imitation voice in the TTS system. Then whether the artist or voice actor agrees to license his or her voice right to the TTS developer is matter of contractual negotiation and the court may not wish to interfere in a case where both parties have similar bargaining power,” Davidson said.

Balancing innovation and fundamental rights

The challenges posed by TTS technology reach far beyond studios and the creative industry, touching the general public’s lives. While prominent artists with fame and influence lead the charge for change, they are backed by a chorus of lesser-known voice actors with less financial ability to take action. Behind them, another group. Every day, citizens have even less power to safeguard their fundamental rights.

AI tech companies offer services like TTS, voice cloning, and voice dubbing, along with features that merge voices with images to create videos that blur the line between reality and fabrication. This raises concerns about the authenticity of what we see and hear, the copyrighted data being used, and the extent of the rights people may unknowingly forfeit when using these applications.

“To the everyday man on the street, I think you still feel like something has been taken from you, or there has been some intrusion into something that is yours or something you think should belong to you,” said Anna Toh, director of Amica Law in Singapore.

When creators take inspiration from existing work, they typically know how to adjust to avoid copyright infringement. However, as Toh explained, “When you’re using AI tools, you don’t know where they are getting their inspiration from. The tool comes out with its own output; I don’t know how much has to be changed to that in order to be further away from the original.”

While AI tools can accelerate mundane tasks, they also open new creative possibilities. These technologies can generate lifelike voiceovers, enabling historical figures to “speak” in documentaries or offering personalized, interactive storytelling experiences that feel like they are coming from a loved one.

However, these advancements also underscore the delicate balance that must be maintained. Haylock points out: “Ultimately, there has to be a balance struck between the protection of individual rights and the pursuit of technological advancements.”