Somewhere in a university server room, a hard drive is spinning its last rotations. On it: an interactive installation from 1994, a piece that once asked gallery visitors to touch a screen and watch the world respond. The artist is still alive. The museum still has the hardware. But the software it runs on hasn't been supported in two decades, and the one technician who understood its architecture retired in 2011. Nobody has opened the files in years. Nobody is sure anyone can.
This is not a dramatic story. No one hacked anything. No one destroyed anything. The work is simply becoming unreachable — not through malice, but through time.
That quiet disappearance is the crisis.
---
To understand why digital art is so fragile, you have to understand what it actually is. A painting is paint on canvas. A sculpture is material arranged in space. But a piece of new media art — an interactive web installation, a video work dependent on a specific codec, a generative piece built in Flash — is a relationship between hardware, software, and human interaction. Remove any one leg of that stool, and the work ceases to exist in any meaningful sense. The hardware becomes obsolete. The software stops being supported. The interaction becomes impossible. What remains is a kind of ghost: files that open to nothing, screens that display error messages where art once lived.
By the late 1990s, a small number of conservators and theorists had begun to recognize this problem with genuine alarm. Jeff Rothenberg put it plainly in his 1998 paper "Avoiding Technological Quicksand: Finding a Viable Technical Foundation for Digital Preservation" — digital materials were being created faster than anyone had developed methods to save them, and the window for intervention was narrowing with every product cycle. Rothenberg wasn't writing about art specifically, but the implications landed hard in museum conservation departments. If documents were at risk, artworks built on the same fragile technological substrate were in even deeper trouble.
The people paying closest attention were a scattered network of conservators, archivists, and curators who had started collecting time-based and digital work in the 1980s and 1990s and were now watching their acquisitions quietly decay. At institutions like the Tate Gallery in London, where Pip Laurenson served as Head of Time-Based Media Conservation, and at the Solomon R. Guggenheim Museum in New York, where Carol Stringari was building one of the most rigorous conservation programs in the field, staff were confronting the same terrifying question: what does it even mean to preserve something that was never meant to last?
---
The first serious attempt to answer that question systematically came in 1998, when Jon Ippolito — working with artists Janet Cohen and Keith Frank — developed what he called the variable media concept. The idea was elegant and slightly radical: instead of treating a work of art as a fixed object to be maintained in perpetuity, you treated it as a set of behaviors and intentions that could be expressed through different means across time. The work wasn't the specific hardware. The work was the idea the hardware had been built to express.
In 2000, Richard Rinehart published "The Straw that Broke the Museum's Back? Collecting and Preserving Digital/Media Art for the Next Century," sharpening the field's sense of its own emergency. Two years later, in 2002, the Guggenheim formalized its commitment by partnering with the Daniel Langlois Foundation — a Montreal-based organization dedicated to art, science, and technology — to form the Variable Media Network. That same year, Timothy Murray founded the Rose Goldsen Archive of New Media Art at Cornell University, naming it after Professor Rose Goldsen, a Cornell communications scholar. The archive became one of the most significant repositories of its kind in North America.
The field was beginning to cohere. Four preservation strategies emerged as the canonical options: storage, which simply meant keeping the original hardware and software intact; migration, which meant moving the work to new formats as old ones became obsolete; emulation, which meant recreating the original technological environment in software so the work could run as intended; and reinterpretation, which meant rebuilding the work from scratch using new tools while honoring the original concept. Each strategy came with tradeoffs. Storage was honest but finite — hardware fails. Migration risked subtle changes to the work. Emulation was technically demanding and expensive. Reinterpretation was considered a last resort, because it fundamentally altered what the work was.
In 2004, the Guggenheim and the Daniel Langlois Foundation staged an exhibition called "Seeing Double: Emulation in Theory and Practice," presenting original works alongside emulated versions and asking audiences — and critics — to evaluate whether the emulations were faithful. It was a public test of the field's most contested strategy. The results were inconclusive in the best possible way: they proved the question was genuinely hard.
---
The anomalies and contradictions in this field don't announce themselves. They accumulate. A conservator discovers that a work acquired in 1997 was built on software that became unsupported in 2003, and that the artist — the only person who fully understood the code — died in 2009. A museum's digital collection turns out to include dozens of pieces that were never properly documented at acquisition, their dependencies unrecorded. A web-based work exists only as a URL that now returns a 404 error, and no one thought to crawl it before the server went dark.
The deeper problem is philosophical. New media art often depends on audience interaction for its meaning — the "liveliness" of a visitor touching a screen, navigating a database, triggering a response. That liveness is, by definition, ephemeral. You can emulate the software. You can reconstruct the hardware environment. But the experience of encountering the work fresh, in a gallery, in 1996, with the specific cultural context of that moment — that cannot be archived. Conservators working in this field have begun to acknowledge openly that some dimension of these works may be fundamentally unrecoverable.
---
The tools that exist are real and genuinely useful. Archivematica, an open-source suite developed for long-term digital preservation, allows institutions to ingest, describe, and store digital objects with rigorous metadata. Conifer — formerly known as Webrecorder — can capture web pages including password-protected content and media that standard crawlers like the Internet Archive's Wayback Machine cannot reach. BitCurator assists with digital forensics, recovering data from aging storage media. PRONOM, maintained by the UK National Archives, catalogs file formats and their histories, helping conservators identify what software a given file requires.
In 2007, Richard Rinehart introduced the Media Art Notation System, known as MANS, which uses XML to create structured documentation of digital artworks — essentially a score, in the musical sense, that describes a work's behaviors and requirements at three levels of increasing technical specificity. The Variable Media Questionnaire, a free web service developed through the Variable Media Network, allows institutions to share preservation strategies and artist interviews. Organizations like DOCAM — the Documentation and Conservation of the Media Arts Heritage alliance, organized by the Daniel Langlois Foundation — and the International Network for the Conservation of Contemporary Art, organized by the Netherlands Institute for Cultural Heritage, have worked to build international standards and shared knowledge across institutions.
Rhizome.org, the digital art platform affiliated with the New Museum of Contemporary Art in New York, has been particularly aggressive in preservation work, developing its own emulation infrastructure and maintaining an archive of born-digital artworks. Franklin Furnace, the performance and avant-garde art organization founded by Martha Wilson, has similarly committed to preserving ephemeral and time-based work.
---
What investigators in this field have confirmed is stark: without early intervention and active collaboration with living artists, the majority of new media works will not be properly preserved. The window for that collaboration closes when artists die, when memories fade, when institutional knowledge walks out the door with a retiring technician. What remains contested is which preservation strategy best honors a work's authenticity — whether an emulation of a 1990s web piece running on a 2024 server is the same work, a faithful copy, or something else entirely. The community has come to believe, with increasing urgency, that the answer depends on the work — that there is no universal solution, only case-by-case judgment calls made under time pressure.
The speculative edge of the field is darker. If a work was built on software that is already obsolete and undocumented, migration may be technically impossible. The source code may be gone. The original developer may be unreachable. In those cases, the only option is reinterpretation — rebuilding from intention rather than from artifact — and whether the result counts as preservation or as something closer to fan fiction remains genuinely unresolved.
Today, institutions like the Berkeley Art Museum and Pacific Film Archive, the San Francisco Museum of Modern Art under conservator Jill Sterret, and Jon Ippolito's Still Water lab at the University of Maine continue to push the field forward. The infrastructure is better than it was in 1998. The awareness is sharper. But the backlog of unpreserved work grows every year, and somewhere, right now, a hard drive is spinning down for the last time — carrying with it something that was once, briefly, alive.