About

Equity is a trade union of 50,000 performing artists and creative practitioners, united in the fight for fair terms and conditions in the workplaces of the performing arts and entertainment industry. Our members are actors, singers, dancers, designers, directors, stage managers, stunt performers, puppeteers, comedians, voice artists, and variety performers. They work on stage, on TV sets, on the catwalk, in film studios, in recording studios, in night clubs and in circus tents.

Introduction

The vast majority of generative AI models in existence today have been trained on creative works without authorisation or remuneration for creators. This is industrial-scale illegal infringement of Equity members’ intellectual property.

The UK government has a legal and ethical responsibility to ensure that the creation and use of AI tools are built on legally compliant data. Generative AI companies must license the performance data that they are using to train foundational models, using the existing, established licensing frameworks. Unethical models built on dirty data cannot become the norm. Enforcement and strengthening of the legal framework will allow the industry to move forward with ethical AI frameworks, delivered via collective bargaining.

Policy recommendations

1. Drop plans for an (opt-out) Text and Data Mining Exception

A Text and Data Mining Exception (TDM) would provide a legal loophole for data mining systems to copy protected works for the purposes of extracting and analysing the data that they contain. It would allow Generative Artificial Intelligence (GAI) companies to use copyright protected works in the development of foundational AI models, without license or remuneration. It is reported that the government is considering introducing a TDM, but with an option for rightsholders to ‘opt-out' from the exception.

This TDM is a significant threat to the creative industries where industrial-scale illegal infringement of Equity members’ intellectual property is already occurring during the development of GAI models. The TDM would throw open centuries of creative content to exploitation without licensing and remuneration, destroying the value of these assets for rightsholders, and undermining long-established property rights.

As Ed Newton-Rex has described, there is no effective way to deliver an opt-out mechanism for rightsholders. Both location-based and unit-based mechanisms are deeply flawed. Neither will rightsholders themselves have the awareness or capacity to enact such a right, were it to be introduced. There are already significant challenges for enforcement of the existing copyright and intellectual property regime upon an AI industry which is highly opaque and currently refuses to adequately acknowledge or protect existing rightsholders. We therefore do not regard the proposal for an ‘opt-out’ as practicable.

Instead, the UK government must require operators of internet crawlers and general-purpose AI models to comply with UK copyright law, even if the training takes place in another jurisdiction. Operators of internet crawlers and general- purpose AI models must be transparent about the identity and purpose of their crawlers, and to be transparent about the copyrighted works they have scraped. Without these obligations, is it impossible to enforce copyright law.

The success of the UK creative industries rests on decades of protected copyright and intellectual property development. A breach of this framework via a TDM could impact inward investment and is highly likely to undermine the growth-potential of our industries, which are now recognised in the

government’s Industrial Strategy 2035 White Paper.

2. Update the performers’ rights framework

The Copyright, Design and Patent ACT (1988) was introduced at a time when the dominant format was the CD. Existing legislation was not designed to effectively regulate the impact of GAI on the copyright and intellectual property regime, and is therefore lagging way behind technological advance. The government states that copyright provisions give performers the “right to control who is able to record and make reproductions of their performances… regardless of the technology used to make such reproductions, including AI technology”. Legal firm Lewis Silkin describes this position as a “red herring” and instead argues that “it is a grey area whether copyright law gives performers the right to control synthetic AI performances which reproduce their voice or likeness”.

We need expressed protection against unauthorised synthetisation because of the narrow legal protection for performances (as compared to authorial works) in the current legal framework:

Section 182(1) should be revised to include performance synthesisation,

and the digital cloning of live performances as an act of ‘recording’;

Section 182A of the Act should be revised to include performance synthesisation, and the digital cloning of recordings as an act of making ‘a copy’;
Alternatively, Part 2 of the Act should be revised to introduce a separate right to control the reproduction of performances.

In response to the Intellectual Property Office consultation on the Beijing Treaty in 2023, Equity outlined its strong support for the introduction of legislation to enact Option 1B which would build upon the existing framework of collective bargaining in our sector, developed over a century and valued by both engagers and the workforce, while strengthening the exclusive rights of performing artists. We believe that the government must provide performers with unwaivable moral rights, to protect them in the face of uneven bargaining power with producers using contract waivers. The government should complete implementation of Option 1B of the Beijing Treaty, but with the addition of making the moral rights it provides unwaivable.

In the UK, there is also no codified law of image rights or privacy. Instead, we have a patchwork of statutory and common law causes of action, which an individual can use to protect various aspects of their image and personality. Advances in AI technology have reinforced the urgent need to introduce "image rights". Provisions of image rights in law enable performers to defend their earnings, artistic integrity, career choices, brand, and reputation. These rights are also an important tool for protecting privacy, and consumer rights. The system of image rights in Guernsey may provide a model for implementation.

3. Apply and enforce the existing UK GDPR regulations

In the absence of a sufficiently strong intellectual property framework, or robust performers’ rights, we are relying on the UK GDPR laws to protect our members from exploitation.

Under UK GDPR regulations, performance data is personal data where it contains any information capable of identifying the performer, such as voice, likeness or movement data. These rules apply where personal data is collected and processed. In these circumstances, data subjects hold significant rights, including around transparency, consent and erasure.

The ongoing development of GAI models which exploit our members’ voices, likenesses or movement without their ongoing, informed consent may breach these rights, in circumstances where our members’ data is being processed as ‘input’ into foundational GAI models, and where they are identifiable in GAI outputs, such as in recent cases where our members’ likenesses have been used to generate ‘deepfakes’ without their consent.

Currently, enforcement is therefore failing to prevent serious breaches in UK GDPR rules during development and implementation of GAI in creative industry contexts. We are calling on government and the ICO to work together to publish clearer guidance on this issue, and for the ICO to use its position as an industry regulator to ensure organisations meet their information rights obligations.

4. Introduce an AI Safety Act to tackle harmful content

The government has introduced significant additional protections for internet users through the Online Safety Act, which places new duties on social media platforms to guarantee a safe environment for users, especially children.

There is significant scope for the government to consider placing similar duties on GAI companies, to ensure that the customers of this technology cannot generate certain types of content, such as content that is illegal, for the purpose of exploiting, harming, or attempting to exploit or harm minors, for making a statement about religion or politics, or in any way that is obscene, defamatory, threatening, fraudulent, bullying, or discriminating.

In 2021, Equity member Dan Dewhirst took up a contract with a company called Synthesia and found that his AI avatar was being used to peddle fake news and propaganda in a highly volatile Venezuelan social media campaign. The lack of regulation and unwillingness of the engager to amend contract clauses left him vulnerable. More about Dan's story can be found on the Equity website.

A set of new duties on the part of GAI developers and user would help to combat misinformation and manipulation of images. This could also provide broader positive guarantees towards a safe and fair democratic process and protecting the image and personality of those in the public domain, including politicians.

For further information, kindly contact Tom Peters, Equity’s Policy Officer.