Slowing down, opening discussion and imagining futures: how we’re developing an AI ethics and policy approach at the Library

At State Library Victoria, we are taking a radically thoughtful, research-driven approach to develop a framework of ethics and policy around responsible engagement with artificial intelligence (AI) 1 technologies. When it comes to AI, State Library Victoria is choosing to not "move fast and break things". Instead, we are doing something novel in our sector. 

Our research looks inward and outward at once. Internally, we are examining the values and principles held by our organisation and staff; drawing upon the deep knowledge and expertise held within the Library to articulate shared values and identify our "leverage points" 2 for action in the wider AI system. We aim to build shared organisational values out of consensus and to translate these into an ethics of AI practice at the Library.  

Externally, we are researching the rapidly changing landscape of AI, policy and governance, considering how changing social, ecological, political and cultural contexts may impact the Library and shape the policy interventions we design. This approach of scanning internally and externally positions the Library as an active participant in technological change, rather than a passive recipient and opens the questions: "How might we innovate with AI technologies at the Library, in ways that enable ‘good futures’ for the State Collection, our publics, communities, staff, and environment?" and "where in the AI space might the Library choose to innovate?". 

AI is a "baggy term" 3, an umbrella concept used as shorthand for multiple technologies with different histories, affordances, applications and potential impacts. Analytic AI 4, computer vision 5 technologies such as optical character recognition (OCR) 6 and forms of natural language processing (NLP) 7 have been in use for decades within libraries, their information systems, and digitisation workflows. But the most recent waves in AI development –  particularly Large Language Model (LLM) 8 based generative AI 9 and multi-agent systems (MAS) 10 pose new possibilities, ethical dilemmas and existential questions for libraries. Modern libraries are information organisations, built around the principles of providing open access to information and knowledge – for all. Libraries had a natural alignment with the early internet and the emergence of the open movement and its values of "universal access to knowledge and culture". 11 Mass digitisation of collections over the last two decades have contributed to our digital commons online, however these acts of enabling access to digitised culture have been exploited in the development of LLMs. We have witnessed the profiteering and extraction of the open web by technology companies who, in the development of proprietary language models, have transformed public access into private ownership.  

Collecting libraries are the custodians and protectors of authorship and original works, whereas for-profit LLMs controversially rely on vast troves of illegitimately accessed authored materials 12 – a point of tension which exposes the vast gap in values and purpose between tech companies and libraries. Generative AI and LLMs promise increased access to vast amounts of cultural knowledge, but they also present threats to the role of librarians as caretakers and research specialists with deep knowledge across complex collections, and risks to libraries’ ability to ensure integrity of information.  

As the AI landscape becomes more complex, this research represents a deliberate decision to slow down and deeply consider the ethical dimensions and policy implications of adopting, implementing and developing AI technologies at the Library. As a state library, a public institution and a caretaker of a state collection, we have a responsibility to take the long view – to consider how the actions we take today, including the development of AI policy instruments, may impact those in the future. Rather than simply react, we have a responsibility to respond to AI with consideration. We recognise our responsibilities to our collection, and its authors and creators, to publics, and communities, to students, researchers and historians, and to future generations of all the above.  

Why ethics and policy research? 

Ever since speaking with Dr Vanessa Bartlett 13 and Jasmin Pfefferkorn 14 in late 2025 about their book, Decentring Ethics: AI art as method 15, I've been thinking about ethics as something that emerges from the frictions, the tensions, the 'rub' – the place where ethics 'happens' and is practiced. In our conversation for the SLV LAB podcast, we discussed how they work with Francisco Varela's concept of ethical 'know how' (ethics-in-action) 16 which sits in contrast to ethical 'know what' (rules, regulation and policy), to explore AI ethics in the cultural sector. Their book is timely, as many of us across the GLAM sector are navigating ethical positions in relation to AI, often in the absence of the ethical ‘know what’ of regulatory measures, legislation and policy.  

It struck me that in this absence of regulation or policy, these ethical dilemmas require us to deeply reflect upon our values and consider what ‘right action’ looks like, in a way that simply ‘being compliant’ or straightforward rule following doesn’t allow. In this moment of ambiguity around AI technologies, we are invited into a larger practice of ethics-in-action. By recognising and engaging with the ethical tensions and frictions we encounter when use of AI technologies challenges our deeply held values, we are actively doing ethics. But to do this at the organisational scale, we first need to know what our shared values are.  

Strategy development typically involves two zones of research or inquiry, perhaps better framed as two questions:

1. Where or what do we, as a collective, community, or organisation, wish to go, or want to do?
2. What is changing in our external environment that might, enable, aid, hamper or prevent this desired trajectory? 

I think of these as the two sides of a coin in research-led organisational strategy. One side invites us to reflect upon the internal landscape – our values, principles and practices. This inward attention lends itself to ethical inquiry: how might collective values translate into ethics-in-action and everyday practice? The other side asks us to cultivate awareness of the external landscape –  to become attuned to what has changed and what continues to change.  

We need to surface our shared values and cultivate ethical ‘know how’ before we can develop our ethical ‘know what’ of rules, regulations and policy. Policy and ethics are intimately linked. Policy documents and tools are tangible crystallisations of an organisation’s values and ethics. On the surface, policy may function as a mechanism for compliance; at a deeper level, it reflects collective consensus – explicit or pre-conscious – about which actions are beneficial and which may cause harm.  

This has been my approach in developing State Library Victoria's program of AI ethics and policy research: surfacing our internal values and an emerging AI ethics held by the Library while scanning our external policy environment. By working this way, we avoid applying generic policy to the broad and multifaceted context of AI at the Library, or producing a set of AI principles that are so wide they become overly open to interpretation. I also feel it’s important not to rush.  For an AI policy, framework or tool to be effective, it needs careful design as well as mechanisms for review, iteration and adaptation as the AI landscape shifts. Equally, we need a strong sense of internal values and ethics in relation to AI to determine what a policy tool should achieve in practice. The stakes are too high: both the risks and opportunities for the Library – and for the publics we serve –  demand careful attention to what is emerging in this complex space, and a clear articulation of how the Library wishes to participate in co-creating an AI future with other actors.  

Our collaborators 

To explore the Library's 'internal landscape' of values, principles and ethics in relation to AI, we are working with Dr Vanessa Bartlett and Dr Jasmin Pfefferkorn. Their recent work champions ‘soft ethics’ as an approach to understanding how artists, museums and GLAM organisations navigate AI in the absence of ‘hard ethics’ such as established regulatory frameworks. 

Vanessa and Jasmin have led a series of AI ethics and policy workshops with State Library Victoria staff and executives. These sessions have , opened a conversational space to explore staff cares and concerns in relation to AI. Alongside workshops, they have been conducting interviews across the Library to identify where ethical frictions and tensions are emerging, as well as where opportunities may exist for AI technologies to support Library practices.  

To explore the Library's 'external landscape' in relation to AI policy and governance, we are working with Elliott Bledsoe 17 as an AI policy researcher. In addition to reviewing AI policy across libraries and the GLAM sector globally, Elliott will work with us to run a Speculative Policy Lab as part of SLV LAB. We will generate speculative design 18 artefacts to research and road-test different AI policy approaches with Library staff and executive, with peers, and potentially with publics.  

Our principles and approach 

We are designing and developing this research process around a set of principles of our own. First, we are slowing down – allocating 12 months for a program of research, discovery and dialogue. Rapid change – such as that currently driven by advances in AI – creates uncertainty and generates anxiety in the system. Anxiety induced ‘fight or flight’ reactions at a systematic or organisational level are of limited value if we wish to formulate a considered response to an external driver of change like AI technologies. A pervasive sense of panic pushes us all into varying degrees of nervous system dysregulation, conditions that are poorly suited to generative thinking, innovation or the ability to recognise and explore pathways for action. 

Second, we have emphasised the importance of dialogue. We've commenced development of an AI policy approach by creating spaces for open conversation among staff. This helps us understand the contexts of those who will ultimately need to use and work within an AI policy or framework, but its deeper purpose is research: to learn how staff are using (or avoiding) AI technologies now, how they may wish to use them in the near future, and what values shape both use and avoidance.  

Third, we are drawing on futures research and speculative design approaches to explore AI policy at the Library. Our Speculative Policy Lab will create artefacts that speculate on AI futures for the Library. Speculative design research brings particular value to policy design by shifting discussion from the abstract to the tangible. Through the Speculative Policy Lab, we aim to create tangible policy artefacts that provoke discussion and deliberation, support further articulation of the Library’s values and ethics in relation to AI, and inform the drafting of policy.  

Continuing SLV LAB’s ethos of sharing knowledge and working in the open, we will publish research updates and share our process. We are developing resources for our peers and peer organisations wrestling with similar questions about how to formulate organisational ethics and policy approaches to AI. In the same spirit, if you’d like to collaborate with us on research, reach out. We are partnering with researchers, organisations, and peers to build a body of research on the future of libraries and emerging technologies, and we will share our first research dispatch later this year. 

Get in touch via our contact form