When Daniel Chávez Heras describes the Intelligent Systems for Screen Archives project, he reaches for a phrase that cuts to the heart of Britain’s most consequential experiment in cultural technology: “Each token bought is capacity not built.” It is a formulation that sounds like academic polemic, but it carries the weight of a sector-wide reckoning. Across the United Kingdom, film and television archives – custodians of over a century of moving image heritage – are confronting a paradox that defines our moment. Artificial intelligence promises to unlock treasures buried in vast, poorly catalogued collections. Yet the tools most readily available come packaged inside proprietary systems that transfer knowledge from the public sphere to the balance sheets of a handful of corporations. The ISSA project, funded by the BFI through National Lottery money, proposes a different path. It is, in Chávez Heras’s framing, “a counter-narrative to Silicon Valley’s AI hegemony.”
The Scale of What Is Hidden
To understand why this matters, one must first grasp the magnitude of what UK screen archives hold and how much of it remains invisible. The collections span the entire history of cinema and television: from nitrate films of the Edwardian era to born-digital works uploaded to YouTube in the past decade. The BFI National Archive alone has recently acquired four hundred online moving image works as part of its Our Screen Heritage programme, charting the explosion of formats, styles, and voices that define the internet age. Regional and national archives across the four UK nations – the National Library of Scotland, the National Library of Wales, Northern Ireland Screen, the North West Film Archive, and Yorkshire Film Archive – hold additional millions of items, many sparsely catalogued, their contents largely unknown even to the institutions that preserve them.
The problem is not merely one of volume. It is one of legibility. Traditional cataloguing relies on human labour: a curator watches a reel, writes a description, assigns subject headings, cross-references related materials. This process, painstaking and skilled, cannot keep pace with the sheer quantity of material, particularly as collections grow to encompass born-digital content whose formats, codecs, and compression schemes proliferate without restraint. The BFI’s own curators have described digital acquisition as “a whole different beast” from the analogue era, where film came in a handful of standard sizes and video production followed predictable engineering standards. Today, a single creator may export a work in any of dozens of configurations, and the archive must verify that each file has a clear preservation roadmap before committing to safeguard it in perpetuity.
What AI Can See That Humans Cannot
This is where artificial intelligence enters, not as a replacement for human judgement but as a radically different mode of perception. The prototypes being developed by King’s Digital Lab under the ISSA banner deploy video language models, semantic search, and retrieval-augmented generation to analyse moving images at a granularity that no team of human watchers could achieve at scale. One prototype breaks video files into meaningful segments and generates descriptive metadata automatically. Another identifies when named places are mentioned or appear visually in a film, enabling searches that would otherwise require exhaustive manual annotation. A third explores audio description – the narration of visual action for visually impaired audiences – as a task that machine learning can support while respecting the artistic and ethical complexity of the work.
Rishi Coupland, the BFI’s Director of Research and Industry Innovation, frames the ambition in terms that transcend mere efficiency: “AI technologies have the potential to unlock enormous potential for screen archives of all scales, however in this fast-moving space we need a much more comprehensive understanding of the opportunities and the challenges facing audiovisual collections.” The ISSA project, he argues, will provide “new tools, skills and insights to establish an R&D framework that could benefit the wider sector in integrating AI technologies in the institutional fabric of moving image archives, while ensuring that we prioritise considerations such as copyright and ethical perspectives.”
The Open-Source Counter-Argument
The decision to build open-source, modular prototypes rather than procure commercial AI services is not incidental to ISSA’s mission. It is the mission. In a presentation to the LUSTRE network in April 2026, Chávez Heras articulated the distinction with precision. When archives purchase AI services – automated transcription, entity extraction, metadata generation – the knowledge of how those outputs were produced remains with the vendor. The archive receives a result but not the understanding. “This transactional model offers efficiency but forecloses institutional learning,” the LUSTRE documentation noted. “Archives become consumers of AI rather than participants in its development, and the sector’s collective understanding of what these technologies can and cannot do remains shallow.”
The alternative ISSA pursues is slower and harder. Five archives across the four UK nations are collaborating with King’s College London to develop shared infrastructure, transferable methods, and crucially, collective knowledge about AI’s capabilities and limitations in archival contexts. The project’s approach alternates between engagement and development: fifteen interviews with archive professionals were distilled into four use cases and technical requirements, followed by iterative prototyping toward minimum viable products, a demonstrator event with all partners, and situated workshops that apply these tools to specific archival challenges. The structure is designed not only to produce functional tools but to generate and circulate practical knowledge across institutions with different scales, capacities, and priorities.
Navigating Copyright and Cultural Sensitivity
The ethical terrain is treacherous. Training AI models on copyrighted films raises legal questions that remain unresolved in UK and European law. Automated tagging risks perpetuating biases – mislabelling cultural rituals, flattening nuance, imposing Anglo-American categories on material that resists them. The ISSA team confronts these challenges through workshops that integrate ethical AI frameworks, addressing questions such as how to honour creators’ rights while enabling computational analysis, and whether algorithms can be taught to recognise culturally specific symbolism without reducing it to a data point.
Annie Shaw, the BFI’s Public Access Researcher and a specialist in copyright and licensing, has been navigating these tensions through the Our Screen Heritage programme. Her work on copyright research, licensing, and risk frameworks for mass digitisation projects illuminates the practical difficulties: archives must balance openness to new forms of moving-image practice with the need to uphold principles of originality, attribution, and public trust in the cultural record. The speed at which AI develops challenges legal frameworks grounded in human authorship and creative processes, creating uncertainty over whether AI-generated works attract copyright protection at all, and who, if anyone, owns them.
Beyond Preservation: A Model for Public AI
The ISSA project arrives at a moment when the question of who controls artificial intelligence has moved from academic debate to political urgency. The European Union’s AI-BRIDGES and ECHOLOT initiatives pursue parallel goals: building interoperable cultural heritage data ecosystems, developing AI-enhanced processing with human oversight, and ensuring that the benefits of machine learning accrue to public institutions rather than being extracted by private platforms. The Archives and Records Association in the United Kingdom has published AI preparedness guidelines through its FLAME project, emphasising that “AI can support archival work, but only when collections are made ‘AI-ready’ through careful preparation, documentation, and governance.”
What distinguishes ISSA is its insistence that the question is not technical but political. Chávez Heras has argued that “we are reaching a critical inflection point in which we have to define the role that AI technologies are going to play in social life, including how we want these technologies to mediate our relationship with over a century of film and television. This is too important to be left to a handful of large companies.” The project’s open-access ethos – its commitment to releasing code, case studies, and best practices as public goods – ensures that smaller archives are not left behind in what could otherwise become a two-tier system: institutions rich enough to buy AI services, and those forced to watch from the margins.
What Success Looks Like After 2027
If ISSA delivers on its thirty-month mandate, completing in August 2027, its impact will ripple far beyond the five partner archives. Regional collections could leverage AI to engage local communities, transforming static repositories into participatory platforms. A schoolchild in Cardiff might explore Welsh mining history through AI-generated interactive timelines. A filmmaker in Belfast could splice archival Troubles footage into virtual reality narratives. A researcher in Manchester might discover forgotten amateur footage of postwar Liverpool, surfaced not by manual input but by machine learning that detects uncatalogued visual motifs and links disparate clips through contextual analysis.
The BFI’s broader innovation strategy supports this vision. The Innovation Challenge Fund, which has distributed up to £1.8 million across successive calls since 2024, treats technological adaptation as existential for arts organisations. The BFI National Archive is simultaneously establishing a £1 million Moving Image Conservation Research Laboratory – the first of its kind in the UK – equipped with multispectral scanning technology capable of capturing the full colour spectrum present in historical films. These parallel investments in material conservation and computational intelligence suggest a sector that understands preservation and innovation as complementary rather than competing imperatives.
A Manifesto for Cultural Sovereignty
In this light, the £192,500 ISSA grant is not a niche academic endeavour. It is a test case for whether public institutions can develop artificial intelligence on their own terms, building capacity rather than renting it, generating knowledge rather than consuming it. The Yorkshire Film Archive curator who participated in the project captured the stakes with an economy that no policy document has matched: “We’re not just saving films. We’re safeguarding the right to reinterpret our past.”
That sentence carries an implication that extends well beyond screen archives. In an era when streaming platforms algorithmically bury older films, when deepfakes muddy historical truth, and when the largest technology companies propose to mediate all human knowledge through their proprietary models, the question of who owns the tools of interpretation is not abstract. It is the question of whether cultural memory remains a public resource or becomes a licensed product. The archives working with ISSA have chosen their side. The rest of the cultural sector is watching.
