Beyond the Ick: Public Domain Day 2026 and the Psychology of Closing the Commons
Tomorrow, January 15th, the Royal Library of Belgium will host the annual European Public Domain Day celebrations. Archivists, librarians, legal scholars, and open culture advocates will gather to mark the moment when works from 1954 enter the public domain across Europe. It should be a moment of pure celebration: another year’s worth of human creativity becoming freely available for everyone to read, remix, and build upon.
But this year, the celebration arrives shadowed by a troubling question that the open community must confront: Are we about to wall off the very commons we fought so hard to create?
The question is no longer hypothetical. Across the internet, institutions that have long championed universal access are now considering, and in some cases implementing, aggressive restrictions on how their collections can be accessed. The stated reason is always the same: protecting against AI scrapers. Major cultural heritage institutions are quietly reviewing their access policies, weighing whether to require logins, deploy bot detection systems, or simply block entire categories of automated access.
The rhetoric is seductive. We’re told this is about protecting creators, preserving the integrity of human culture, and preventing exploitation by Big Tech. But if we’re honest with ourselves, something else is driving this impulse, something more visceral and less rational. We need to name it: this is about disgust. About the “ick factor” of AI-generated content. About a psychological reaction that, however understandable, threatens to undo decades of work toward a truly open internet.
The Uncanny Valley Comes for the Commons
In 1970, robotics professor Masahiro Mori described what he called the “uncanny valley“: that eerie sense of revulsion we feel when something almost, but not quite, resembles a human being. The more realistic the approximation, the more disturbing the small imperfections become. Dead eyes. Unnatural movement. The subtle wrongness that triggers an instinctive recoil.
This phenomenon has migrated from robotics into our relationship with generative AI. We encounter it constantly now: the LLM that writes with perfect grammar but occasional logical impossibilities. The image generator that produces stunning landscapes with anatomically impossible trees. The chatbot that mimics empathy while lacking any actual understanding. Each interaction reinforces a growing sense of unease. This isn’t quite right. This isn’t quite human. This is… creepy.
Recent research confirms what many of us feel intuitively. Studies on AI-generated text show that readers experience unease and creepiness when encountering content that is fluent but emotionally flat, confident but factually unstable. The term “slop” has entered our vocabulary to describe this AI-generated detritus flooding the internet. Low-quality articles written by bots. Derivative images clogging search results. Comments sections filled with synthetic engagement. The word itself carries the disgust we feel: slop. Something to be disposed of, not consumed.
And this disgust, however justified it may feel in the moment, is now influencing policy decisions about the fundamental architecture of the internet.
The Paradox We’re Creating
Here’s where things get philosophically thorny. The open community has spent decades arguing for a principle that seemed self-evident: if you have lawful access to read something, you should have the right to analyse it computationally. This was the foundation of the text and data mining movement. “The right to read is the right to mine,” as Peter Murray-Rust and the Open Knowledge Foundation put it in 2012. This principle helped establish legal protections for researchers who wanted to process vast corpora of scientific literature to identify patterns, cure diseases, and generally understand human knowledge more efficiently.
The 2019 EU Copyright in the Digital Single Market Directive codified this through Articles 3 and 4, creating exceptions for text and data mining. These provisions recognised that computational analysis of lawfully accessed content serves the public good. They understood that the ability to “read” with machines was becoming as fundamental as the ability to read with eyes.
But now, faced with AI training, many in the open community are inverting this principle. The argument, often unstated but clearly implied, is becoming: “The right to read is for humans only.”
The distinction being drawn is moral rather than technical. “Good” mining (academic research, nonprofit analysis) remains acceptable. “Bad” mining (commercial AI training, Big Tech data harvesting) must be blocked. Some stakeholders are implementing technical barriers to prevent the very computational access they once fought to legalise, justified by the belief that some actors are exploiting rather than analysing, extracting rather than studying.
This echoes a long-standing tension in the open source software community between “Freedom Zero” and “Ethical Source” licensing. The traditional Open Source Definition explicitly prohibits discrimination against fields of endeavor. You cannot stop the military, oil companies, or surveillance capitalists from using open source software. The freedom is absolute and universal, or it’s not freedom at all.
But the Ethical Source movement argues for restricting usage based on purpose: banning use for surveillance, harmful AI, or other applications deemed unethical. It’s an understandable impulse. Why should our work empower those who would use it for harm?
The problem is that once you accept purpose-based restrictions, you’ve fundamentally changed what “open” means. It’s no longer universal access; it’s conditional access. It’s no longer a commons; it’s a gated community where entry requires proving your intentions are pure.
Freedom of Speech for Knowledge
This is where an uncomfortable analogy becomes unavoidable. If we truly believe in the right to access knowledge, we must think about it the way we think about freedom of speech. The principle of free speech doesn’t exist to protect popular, agreeable, or “good” speech. It exists precisely to protect the speech we find objectionable, because any mechanism that can silence speech we dislike can be turned against speech we value.
Similarly, genuine open access means information is available to everyone, including actors we dislike. Including Big Tech. Including AI companies. Including commercial entities that will profit from what they find. This isn’t because we approve of how they use the information. It’s because any mechanism we create to exclude them can and will be used to exclude others.
When we build systems to block “Bad Tech,” we inevitably create surveillance infrastructure that undermines the privacy and anonymity that made the open web powerful in the first place. Who are you? Are you a bot? Prove you’re human. Show us your credentials. Tell us what you’ll do with this information. These are the gatekeeping questions we spent decades trying to eliminate from public knowledge access.
If Reddit can block the Internet Archive to “protect” its community data, can a government block foreign journalists? Can a library block researchers from countries we’ve decided are “hostile”? The problem is that experience shows that the tools of exclusion, once forged, don’t respect the boundaries we imagine for them.
The Privatisation of Truth
This brings us to perhaps the most troubling consequence of walling off the commons: we’re accidentally creating a data oligopoly where only the wealthiest actors can afford access to truth.
Reddit’s $60 million annual licensing deal with Google illustrates the dynamic perfectly. By blocking open crawling while selling access to select partners, Reddit has created a two-tier internet: one where paying customers get unrestricted access to human-generated conversation, and everyone else gets locked out. This isn’t protecting the commons. This is privatising it.
The irony is devastating. In our effort to resist Big Tech’s exploitation of public knowledge, we’re ensuring that only Big Tech can afford the data needed to build capable AI systems. Open source AI projects and academic researchers will be forced to train on inferior datasets, synthetic data, or whatever scraps remain accessible. Meanwhile, Google, Microsoft, and OpenAI will negotiate exclusive licensing deals for the richest veins of human knowledge.
The result will be AI systems that reflect corporate values, trained on corporate-approved data, optimised for corporate purposes. The messy, democratic, genuinely public knowledge (the arguments on Reddit, the technical discussions on Stack Overflow, the cultural memory preserved in archives) will remain locked away, “protected” from the machines but invisible to the future.
And this phenomenon is not a hypothetical possibility. It’s already happening. Every institution that closes access to “protect” against AI is making the same calculation: better to keep this knowledge locked away than risk it being used by the wrong people. But locked away from whom? Not from those who can pay. Only from those who can’t.
What the Cultural Heritage Community Must Reckon With
As tomorrow’s Public Domain Day celebrations approach, those of us who care about open culture face an uncomfortable truth: the “Great Enclosure” of the internet hasn’t happened yet, but we’re actively considering it. The templates are being drafted. The policy proposals are circulating. The technical solutions are being designed.
We need to be honest about what’s driving this impulse. It’s not purely about creator rights or fair compensation, though those are legitimate concerns. It’s not purely about preventing exploitation, though that matters too. At its core, this is about disgust. About the visceral reaction we have to AI “slop.” About the “Uncanny Valley” we experience when machines produce something that looks like human creativity but feels fundamentally wrong.
That disgust is real. It’s valid as an emotional response. But it’s a terrible foundation for policy.
Consider what we’re actually proposing. To keep machines from accessing public domain works, we would:
- Require login systems that track who accesses what, destroying the anonymity that protects vulnerable researchers, activists, and citizens
- Implement bot detection that inevitably blocks accessibility tools, archival systems, and research infrastructure alongside AI crawlers
- Create a precedent that “public” access really means “approved” access, subject to judgment about purpose and intent
- Establish technological and legal frameworks that can be repurposed by any institution that wants to control how their “public” collections are used.
However, these are not protections. They are mechanisms of control.
A Different Path Forward
There are genuine problems we need to address. AI companies are profiting from publicly available data without contributing back to the commons that sustains them. The flood of AI-generated content is degrading the quality of online spaces.
But the solution isn’t to abandon openness. It’s to demand something better.
Instead of blocking access, we should demand reciprocity. If AI companies train on the commons, they should contribute back: financially through something like a “commons tax” that funds the institutions preserving public knowledge, or through data contributions that make their training datasets and methodologies transparent and accessible.
Instead of implementing crude blocklists, we need machine-readable rights expression. Standards like C2PA or Creative Commons’ work on signaling could let creators express preferences without breaking the web’s infrastructure. The goal should be information, not obstruction: making it clear what uses are encouraged while keeping the content accessible for those who wish to respect creators’ intentions.
Instead of treating all computational access as suspect, we should preserve the right to read anonymously. This includes reading with machines. Researchers, archivists, accessibility tools, and yes, even AI systems, should be able to access genuinely public information without surveillance. The alternative is a web where every access requires authentication, every query is logged, and every use is monitored.
Most importantly, we need to resist the urge to let disgust dictate architecture. The “ick” we feel toward AI slop is understandable. But building the future of knowledge access based on revulsion toward a particular technology is shortsighted. Technologies change. Uses evolve. The principles we establish now will outlast our current concerns.
The Choice Before Us
Tomorrow’s celebration marks the end of decades of enclosure for thousands of works. But the public domain is not just about expired copyrights; it is about the radical principle that knowledge belongs to everyone. Not just the “worthy.” Everyone. That principle only holds weight if we defend it even when the beneficiaries, like AI companies, make us uncomfortable or annoy us.
We are at a decision point. We can let the psychological discomfort of the “Uncanny Valley” drive us toward a new enclosure, erecting the very surveillance and gates we once fought against. Or we can choose to cross it. The way out is not retreat; it is acknowledging the “ick” without sacrificing the architecture of the open web.
Once we accept that openness is conditional and that the commons as a result, requires a background check, we lose something irretrievable: the possibility of genuine universality. The works will survive behind walls, but the dream of an open web will not. Tomorrow, we celebrate that dream. The question is whether we still believe in it enough to save it. I know I do.
Get the article’s core concepts in a unique auditory summary—listen to the AI-generated track below!
This song was created entirely using artificial intelligence tools. If you enjoy this experiment, keep an eye out for each song illustrating our December articles.
Written by Caroline De Cock, LL.M., Head of Research
