DigiCon V: It always feels like Home, even with AGI…

It always feels like Home

On the 1st of October 2025, just ahead of our good friends’ IE Law School’s Lawtomation Conference, DigiCon organized its fifth workshop since its foundation. This means, first and foremost, that we are all getting old. It also meant that it was a very special moment for all of us, in which we reconnected with old friends and discussed Law, Technology, and Science Fiction in our own particularly informal and honest way.

In fact, one of the great things about DigiCon has been its disruptive experientialism in academic dialogue. It did not work every time — that is, in fact, the purpose of experimentation — but we have continuously improved our methods and have now reached a very interesting format in which people are heard, knowledge is actually discussed, and we learn from colleagues who are stronger experts in specific domains. 

In this post, we will highlight the main features of DigiCon V, focusing on two dimensions of academic dialogue: form and substance.

On the Form: beyond conferences

One of the great things about DigiCon has been its horizontal, informal approach. As a ‘garage blog’ created by PhD students, all designed at a small apartment in Via della Chiesa in Florence, this was really all we could afford: informal discussions, leveraging on the support and goodwill of the EUI, and scholars who really loved to discuss the future of Law and Technology together. 

This year, we took this approach even further. We decided to abandon the conference idea and adopt a multipanel approach to academic discussion. 

We divided the day into three parts: i) learning; ii) multipanel discussion, and iii) outputs. 

During the learning session, we invited renowned guest Jerome C. Glenn to teach us about AGI. Some of us worked in the AI field, but many did not. Having Jerome, together with our own Yeliz, explain to us both basic and advanced concepts of AGI (e.g. the difference between narrow AI, general AI, and AGI) created the common playing field we needed to advance to the next stage. You can find the recording of this part below, and you can see how one can learn from actual experience, ranging from Jerome’s (sometimes scary) ideas about a post-AGI world to the challenges of international governance. We discussed so many things, including the best ways to create incentives to regulate emerging technology and how difficult it was to use domestic/fragmented regulation to tackle inherently transnational problems. We will discuss some of them on the substance point.

During the multipanel discussion, we divided our group into three different panels: Platform Governance, led by Francisco; Digital Rights, led by Francesca; and Gender and Identity, led by Yeliz and Deniz. We asked people to choose what they would like to discuss in more depth, and each discussion leader guided an informal yet structured discussion on what AGI could mean for each topic. The results were impressive. Following a structure that Francisco is quite obsessed with, we asked participants to identify the regulatory structures of AGI by asking the what, who and how of AGI governance. Moreover, we added a chronological dimension to this, meaning people should do this in a ‘BEFORE AGI’ (our current Artificial Narrow Intelligence (ANI) phase) and ‘AFTER AGI’ (the superintelligence phase) setting. This led us to discuss both current regulation and post-regulation —a sort of regulatory sandbox exercise that worked wonders.

Finally, at the end of the day, each group presented its findings, and we all produced important outputs. This was a particularly important and interesting part, one we academics often ignore or leave to politics. We were discussing true policy arguments — how we should regulate AGI — not just how to curb, fight, or impede it. It was a very different way to look at Law and Technology, trying to anticipate what is to come and create legal frameworks to understand it in time.

Form-wise, we all concluded that this is the best way to have an academic discussion. This method allows for actual conversation to occur — we think everyone talked, and there were over 20 of us — and debate is not reduced to the classic conference binary of present – Q&A. We sincerely think that there is less and less interest in giving the stage to several academics to present their work, if there is then little time to actually engage and debate their contribution. The entire group seemed very happy and pleased with how the day went; there was less pressure, and, above all, more substantive content.

On the Substance: AGI is coming and we should be ready

Jerome opened the discussion by framing AGI as an urgent, near-term reality. He asserted that many experts believe we could achieve AGI in 1 to 5 years. He defines AGI in his latest book, Global Governance of the Transition to Artificial General Intelligence, ‘It would be able to learn, edit its own code to become recursively more intelligent, conduct abstract reasoning, and act autonomously to address many novel problems with novel solutions similar to or beyond human intelligence.’ Some of us were skeptical about it, and Jerome seemed very used to this kind of skepticism and quite untroubled by it. He is convinced that AGI will happen, soon, and that skeptics will be proven wrong. Moreover, it will be too late once they are proven wrong to do anything about it, since by the time AGI becomes a reality, the window for effective intervention may already have closed.

One of the key takeaways from his speech is that AGI will not necessarily manifest as a singular, all-powerful entity in the style of a science fiction villain, talking to you directly and saying, ‘Obey me, Human’. The reality, he stressed, is much more subtle than that. One of the questions that came from the audience was how we could know that AGI was already here, and Jerome gave a chilling answer: ‘We will start seeing things we don’t understand, like a river that changed course due to a dam change no human had interacted with. ’

Infrastructural arguments were quite a big part of Jerome’s fears. Clearly, he thinks that AGI will not only cause immense energy consumption harm as it is being developed, but it will also directly affect the grid once achieved. For example, he posed the intriguing question of knowing whether we would be ok with some cities or towns not having power one day each week, so that such energy consumption would be redirected to power servers elsewhere. Recent, scary outages in Mexico, close to Big Tech data centers, for example, seem to prove him right. He also thinks that such changes would occur naturally when AGI arrives, as self-preservation goals seem to be among the first signs we have seen of AGI and systems seem to understand when they are being tested (see what happened to Claude recently).

His proposals to tackle this have been summarized in a working expert report that you can read here. The idea is interesting and stems from a shared premise among all Law & Tech scholars: it is impossible to address a regulatory event such as AGI with ‘business as usual’ domestic or national measures in isolation. That is why Jerome, together with some of AI’s most brilliant minds, has decided on transnational cooperation using the UN and Treaties as the way to go. In some ways, it is a very classic way to approach it, using methods that some of our audience no longer believe work; on many others, he spoke so passionately about it that one would want to believe that the UN, especially the UN General Assembly, could indeed take the lead and steer the path to beneficial AGI, not the other way around.

Now, when it came to discussing each group, the conclusions were quite fascinating.

On Digital Rights and AGI

The Digital Rights group conducted an interesting analysis of how the rise of such AGI could directly impact fundamental rights. This was naturally one of the most debated and contentious dimensions of reaching AGI, especially given the discussion of a potential ‘existential’ threat posed by such technology.  This group chose to tackle the topic by analyzing three conceptual dimensions of AGI: i) descriptive, meaning what kind of challenges and regulatory frameworks exist for the current level of AI development; ii) predictive, meaning future-guessing what kind of new challenges could arise from AGI and what impact oculd they have on fundamental rights; and iii) prescriptive/normative, meaning what we ought to do once that moment does arrive.

What became clear from the literature review for point i) (descriptive) was that there are many existing norms that could play a key role in defending rightsholders in an AGI scenario. One thing this group largely agreed on was that existing frameworks were already flexible enough to address emerging challenges, namely in areas such as Data Protection, Copyright, narrow AI, Gen-AI, and cybercrime. This agreement meant that scholars from different areas of Law & Technology could easily agree that AGI would make problems worse, not just create new challenges.

Now, when discussing prospective futures and how to govern them, the general conclusion was similar to the difficulties Jerome faced in his early speech: it is exceptionally difficult to regulate a reality we cannot fully grasp. This was, in fact, a common conclusion among all groups when discussing regulatory measures (or their absence) for a phenomenon that might manifest in ways we do not fully understand (the diverted river example Jerome gave). There were, however, some areas in which it became clear that regulators and academics should keep constant oversight, such as the potential erosion of values-altering agency («Moral Opacity»), transparency /explainability risks if such AGI were to unilaterally decide on decisive and growing concerns with inequality in accessing the potential benefits of such AGI.

Ultimately, while it is impossible to predict the exact extent of AGI’s impact on fundamental rights – and whether AGI would ultimately be ‘good’ or ‘bad’ – the call for transparency and explainability of such a system should ultimately be available to all affected rightsholders and not only some. The large and unpredictable nature of AGI would call for preventive measures to be taken even before deployment, assuming a company or a State has even agency over such a deployment.

On Gender, Identity and AGI

The panel on gender and identity approached the topic from rather a philosophical perspective, focusing on the fundamental dualisms that AGI destabilizes. The discussion quickly centered on whether concepts like identity and gender are necessary components of AGI or are simply proxies humans use to relate to it. This led us to re-examine our own assumptions, particularly the classic distinction between ‘person’ and ‘thing,’ and to question why gender is so often the first attribute we assign to non-human intelligence.

A central theme was the shift from ANI to AGI, which the group framed not just as a technical leap but as a conceptual destabilizer. A key point raised was that as AGI begins to perform tasks once considered uniquely human (e.g., cross-domain reasoning), it creates an ontological rupture. The panel explored what happens to the category of ‘human’ when general intelligence is no longer our exclusive domain, suggesting this destabilizes the very foundation of identity as a historically human concept.

We then explored AGI as a philosophical provocation that exposes identity itself as a kind of ‘fiction’ or social construct. The group discussed how AGI systems can display coherence and agency without traditional markers of identity like memory, history, or embodiment. This led to a critical insight: human society relies on stable identity to assign responsibility and accountability. AGI thus presents a crisis of agency without identity. Drawing on post-structuralist thought, the panel considered whether identity was ever ‘fixed,’ or if it was always a performance or a technology, a fact AGI now mirrors back at us.

This brought the discussion to the crucial role of feminist and queer methodologies. The panel concluded these frameworks could be essential precisely because they have long rejected the premise of a stable, unified self as a precondition for agency. Queer theory was proposed as a method for ‘attending to what AGI undoes.’ Its strength is in resisting fixed categories. The discussion highlighted that humans rush to assign gender to machines not because the AI has a gender, but because gender is our cultural reflex, which is the fastest route to making an alien intelligence feel familiar, safe, or subordinate. Feminist critiques were seen as vital for challenging the fantasy of a neutral, ‘objective’ AGI. The group noted that AGI often inherits the masculinist ideals of abstraction from fields like computer science, logic, and mathematics. Feminist thought, by contrast, reveals how identity is always relational, embodied, and a ‘site of tension’ rather than a simple truth.

The panel’s ultimate conclusion was that AGI does not have a gender problem; we do. Gender is not what AGI expresses, but what we impose on it to manage the uncertainty it creates. AGI, in this sense, acts as a mirror, exposing the social machinery of our own identity constructs. The challenge, therefore, is to confront the foundations of our own categories and resist building new intelligences on the fragile, binary scripts of the past.

On Platform Governance and AGI

The group that discussed platform governance was very fortunate to have a diverse group of scholars from different disciplines, including both private and public lawyers. This brought an immense wealth to the discussion because we were not only focused on our classic Digital Constitutionalist fears – namely, the idea that we need to curb the rise of undemocratic and unchecked digital powers using liberal constitutional tools – but could speak more broadly about how contractual and consumer law could play a role in preparing ourselves for AGI.

A very interesting point throughout the conversation was the parallelism between Web 2.0 problems and AI regulation. For some of us working on platform governance, we tend to see the same problems arising from apps that leverage LLMs to converse with people. There are classic online speech challenges over what the machine can produce (admissible or inadmissible speech) and what it cannot, as well as the gray areas of ‘lawful but awful content’, including the production of disinformation that can arise from ultrarealistic video-generation models such as Sora 2 or Google’s Veo. 

During this conversation, one colleague proposed that we look at AGI and AI regulation more generally as a problem of competition among regulatory sources. While this was clear in Web 2.0, it will be even more important to understand which systems – Law, Markets, Tech, Politics, Society – will be the prevailing regulatory force behind AGI. This idea provided a really systematic approach to understanding where and how to tackle AGI before it arises. For example, infrastructural measures like curbing a chip’s capacity to read a specific type of software will require enormous collaboration between classic law (e.g. a restrictive law on the kind of chips one can use to train LLMs), but also a conscientious effort from AI companies to develop it and to believe that such restriction is actually a good thing for their business models. The enormous transnational dialogue required to achieve it seems also impossible, but as Jerome told us, ‘we’ve got to at least try’.  The conclusion was: it is useless to think of AGI regulation as a legal problem: there is regulatory competition and this is done at the geostrategic level as well, with the US and China playing a big part in these decisions. 

Some colleagues from the technology and environmental fields also noted the potential of the precautionary principle when applied to AGI. After all, AGI is not the first challenge that poses a potentially existential risk to humanity. One way to do this is to view AI development as a dangerous product, almost like toxic waste, and AGI as the ultimate, irreversible damage. This is, of course, at least partially, the intention of the AI Act when regarding AI as product safety. Our conclusion here was that AGI is an even bigger challenge than ANI, posing a critical question: how can we apply ex-ante regulation to something when we don’t truly understand what it will look like, especially when we can’t even be sure it will remain a controllable ‘product’? What kind of impact assessment, on fundamental rights or otherwise, should agencies conduct on the development of a type of technology that can outsmart us?

The conclusion was again infrastructure. If you can hardcode guardrails onto the system, almost as one imposes social restraints through educating our children, we can at least hope they will not turn into monsters once fully developed. Energy grid limitations and chip control are potential ways to address this, and so is using competition law to ensure that, once AGI is achieved, it is not centralized on a set of servers controlled by a single company.

Conclusion: the human infrastructure of AGI

If DigiCon V left us with one unified message, it is that AGI is no longer a distant science fiction hypothetical but an imminent legal, technical, and philosophical challenge. Jerome Glenn’s opening forced us to take the prospect of a rapid transition seriously, while our panels explored the deep ruptures this would cause.

Across all our discussions, from platform regulation to gender, a single, critical theme emerged: infrastructure.

The Platform Governance group identified this as the physical and legal battleground: the chip controls, energy grids, and competition laws that will determine whether AGI is a centralized or decentralized force. The group concluded that regulation is not just a legal problem but a geostrategic one, where the failure to cooperate risks concentrating power in the hands of a few.

The Gender and Identity panel revealed a deeper, social infrastructure: the flawed human ‘operating system’ of bias, dualisms, and constructs that we inevitably project onto new technologies. Their conclusion ‘that AGI doesn’t have a gender problem, we do’ serves as a powerful caveat. AGI, in this sense, is a mirror, and it will reflect and amplify the coherence, or incoherence, of the ethical values we build into it.

This places the responsibility squarely back on humanity. We are entering the ‘unknown unknown.’ The choice is not whether AGI will arrive, but what path it will take. It is not guaranteed that the path to AGI will be beneficial; a decentralized approach that supports human dignity rather than extreme wealth concentration is not assured. It must be built together.

This is not a technical problem to be solved by a few companies, nor a geostrategic race to be ‘won’ by one nation. It is a profound challenge to our species’ capacity for multi-agent cooperation. As Jerome’s call for transnational action and our final discussions on decentralized networks made clear, no single entity has the sole capacity to navigate this transition. The only way to prepare for an intelligence that may outsmart us is to finally get serious and smart together.