Second AI and Creativity Workshop: A collective approach to AI and copyright
On Tuesday 20 May, the second AI and Creativity workshop took place at Westminster, and brought together representatives from the technology sector, creative arts, academia, and government to discuss the impact of AI on the creative industry. The event was organised by the Accelerator Fellowship Programme of the Institute for Ethics in AI, University of Oxford, led by Dr Caroline Green, Director of Research at the Institute, and hosted by Baroness Beeban Kidron.
The workshop explored collective solutions that will enable both creator and AI companies to work collaboratively, respecting creator’s rights whilst supporting technological development. Provenance, consent, education and legislation were some of the topics discussed. The conversation also emphasised the need for developing the UK into a global ethical AI development hub.
Provenance matters both for creatives and AI companies
Professor John Collomosse, from the University of Surrey, held the first session of the workshop, during which he delved into the usability of provenance frameworks, watermarking and metadata. With AI systems relying on data to develop, Professor Collomosse explained the benefits of provenance frameworks, which showcase the origin of a specific creative piece of content, (e.g. photography). Such frameworks allow creators to have better control over their work, enabling them to include information about where their work comes from and how it can be used by third-parties, including AI systems. They also help platforms and consumer identity authentic content, whilst also enabling AI training through opt-in/opt-out mechanism for datasets. The C2PA standard, a cross-industry initiative co-developed by Adobe, the BBC, Google, Meta, Microsoft and others, Adobe’s free Content Authenticity app (Content Credentials), Content ARCS and JPEG Trust were a few of the provenance frameworks and tools that were mentioned currently available for creatives.
Professor Collomosse emphasised the need for making these credentials more durable, potentially by combining watermarking and fingerprinting techniques. However, a significant limitation was acknowledged - watermarks are always removable or strippable, and as there are no industry standards to counteract this, there is nothing to stop this.
The growth in data used to train generative AI models was raised as a concern. According to participants, large quantities of copyrighted, scraped and unlicensed content are already being used to train generative models, often without permission or compensation. It was particularly noted that synthetic data can degrade foundational model training, and some datasets may be "poisoned" or contain illegal, unlicensed content. Therefore, knowing the provenance of AI training data is crucial for addressing risks in the generative AI content supply chain.
The need for machine-readable, open standards for opt-in/opt-out mechanisms for data usage was also stressed, with C2PA offering capabilities in this area.
Scrape bots and transparency
The question about how data is being collected to be used for AI systems was answered by Lucky Gunasekara, Miso CEO. Gunasekara explained in the second session of the workshop that over 16,000 active web crawlers are now scraping news, academic and media websites on a daily basis. While some publishers like The Guardian are using bot-blockers, some of the web crawlers now are mimicking human scrapping patterns, making it even more difficult to block. Data brokering and scraping services are now growing, making traceability of data even more difficult. At this rate, Gunasekara is predicting a significant decrease in referral traffic for publishers within the next few years. A potential solution could be new friction-based techniques for blocking automated scraping, according to the Miso CEO.
Participants in the workshop from the music industry also raised their concerns. As AI systems cannot function without absorbing vast amounts of creative content, it was emphasised that content should not be used without permission. Looking ahead, some of the companies working in the music industry are now seeking to use AI to enhance artist creativity and generate new income, partnering with ethical AI model providers for artist compensation.
The urgent need for transparency was stressed, and suggestions such as labelling of AI-generated content, and full disclosure of training data sources surfaced. Accountability and liability in this new landscape were also suggested as issues that need to be addressed.
Limitations and Exceptions to Copyright
Chris Morrison, Head of Copyright and Licensing at the Bodleian Libraries, University of Oxford, emphasised that limitations and exceptions to copyright are essential elements of the legal framework. He highlighted their importance for facilitating research and educational uses of AI, underscoring a need to maintain a balanced approach beyond licensing solutions alone. Professor Phil Torr from the University of Oxford echoed the significance of exceptions in supporting AI research.
The Importance of Cultural Institutions
Sir Drummond Bone, Byron Scholar and Master of Balliol College at the University of Oxford until April 2018, highlighted the role of collecting organisations, specifically libraries, museums, and archives, as essential in the ethical curation and dissemination of data for AI. Their critical role in preserving cultural heritage while supporting AI research was acknowledged.
UK AI sovereignty
Professor Phil Torr, from the University of Oxford, expanded the conversation from the domestic market to the global market, focussing on UK’s current position in the AI global market. Professor Torr highlighted that the UK must not become an AI colony, supported by American or Chinese AI infrastructure, which are currently front runners in this industry. He advocated that the UK should develop its own AI infrastructure by investing in national datasets, and legislation that supports a fair, safe, and sovereign environment for AI development.
A collective approach
A policy-focussed discussion was held in the last session of the workshop. There was broad recognition that public understanding, clarity in terms of regulation and technical standards must develop together with support from various institutions. The need to build trust and educate the public about AI was highlighted, and the idea of a national campaign was proposed. Calls were also made for proportionate transparency in record-keeping and disclosure, tools for creators to take back control of their widely disseminated work, and educational frameworks for businesses.
Furthermore, the debate acknowledged that AI companies are not against transparency but seek viable licensing models.
The sentiment was that the UK needs to be "loud and proud" in seeking solutions for "sovereign/responsible AI."
Some of the suggestions included:
- Making metadata stripping illegal.
- Requiring AI developers to keep auditable training-data logs.
- Building frictionless licensing systems for creators.
- Launching public-facing campaigns to educate on topics such as AI and copyright.
- Forming a cross-sector “AI and Creativity Alliance”.
- Exploring collective action, dubbed a “NATO for news”, to defend publisher interests.
Baroness Kidron concluded there is "no silver bullet" for AI copyright regulation. She said: “This is not just about AI and copyright. It is about what kind of creative economy, what kind of society, and what kind of digital future we want. The good news is we have brilliant minds working on this, and the responsibility to act together.”
The workshop concluded with the idea of a cross–sector AI and Creativity Alliance to continue these conversations, backed by public awareness efforts and input from creators, technologists, educators and policymakers.
Looking Ahead
Dr Caroline Green said: “The speed of development of AI means we need fast and joint efforts to find solutions for protecting what we value in human creativity while enabling responsible AI innovation”.
We encourage those working in this space to join our steering group, which will help shape future activities and priorities emerging from this work. You can register your interest by completing this short form: https://forms.office.com/e/tQmEb65g6y
Responsible AI requires coordinated action across sectors, borders, and political lines.
