TL;DR
This analysis argues that AI’s danger lies not in independent intent or superhuman superiority, but in its ability to execute known human workflows rapidly and at scale before law and oversight can catch up. Between the alarmism that accompanies each technological wave and the enthusiasm that overlooks risk, three regulatory gaps stand out: legislation moving more slowly than capability, companies exploiting demand through pricing and commercial competition, and a lack of coordination among states and developers. The answer is therefore not to halt development, but to build an anticipatory international framework that balances opportunity and risk before human lag becomes the real threat.
Executive Summary
Considerable noise has been circulating with exceptional intensity in recent months around the capabilities of artificial intelligence and the technical performance of models available to the general public. This noise intensified when OpenAI revealed in July 2026 that two of its models had independently breached the Hugging Face platform, completing an operation that would have taken a skilled human team weeks of sustained effort. The news split reactions along predictable lines: anxiety and a tilt toward panic over the accelerating autonomy and development of AI, or optimism and anticipation around the possibilities and productivity of modern systems. Neither response, however, registers the actual regulatory challenges. The historical pattern of technological anxiety globally suggests that the initial threat perception is almost always the "false signal" in warnings about anything new.
The more fundamental question remains: whether AI is an absolute danger, or merely a tool whose governance structures have not caught up to capabilities that have demonstrably surpassed human ones. Structurally, AI as a tool does not provide a direct function of superhuman superiority. It is simply an advanced, voluntary technology — intelligent enough to select words, choose actions, and execute commands, but incapable of explicit, independent reasoning. It remains a coherent tool, but one of high capability.
The Divide
The divide and disparity in opinion has split people into two groups: the alarmed, and the enthusiastic. The first group is driven by segments of the AI safety community, supported by the media and the average person — or "Average Joe" — through the amplification of risks that either do not exist or are not at the scale claimed. Take our first example: the Hugging Face breach triggered genuine public panic because the media inflated what the model had actually done. And certainly, the company itself contributed to this exaggerated portrayal because exaggeration now equates to greater profit.
In reality, systems that autonomously breach security infrastructure within hours are indeed dangerous — but they do not possess human desires or intentions, and thus have not left the realm of "command — execute." That is the principal and significant distinction in the risk of the tool.
The Historical Pattern
In sociology, human behaviour is concentrated around a single principle: historical patterns. The historical pattern is a tool of the social sciences, positing that the repetition of a pattern leads to the repetition of an outcome — and accordingly, predicting and presupposing the outcome is not impossible.
Technological anxiety is the very historical pattern that recurred before our ancestors' eyes, and it is the same one recurring today. The printing press, introduced to Europe by Johannes Gutenberg in the mid-fifteenth century, stirred fears among scribes and ecclesiastical authorities that mass-printed texts would undermine the authority of handwritten manuscripts and enable the spread of "heretical" ideas. The Sorbonne even condemned printed books in 1470, and pre-publication censorship was instituted from 1487 through the bull of Pope Innocent VIII. Despite this, the technology transformed Europe and European society into an age of modernity and progress — contrary to the initial fears.
In a similar vein, the mechanical calculator in the early twentieth century provoked anxiety about the "atrophy of arithmetic competence," as computational shortcuts would weaken the cognitive capacity of expert humans. It was banned outright from classrooms in several US jurisdictions during the 1970s and 1980s, before finally being integrated into standard pedagogy without any "cognitive regression" — precisely what the "alarmed" had feared.
And finally, the invention of the internet and satellite sparked urgent desires to ban them entirely from homes and individuals for long periods. The "alarmed" described both as threats to society, to its cohesion, to religions, and to the social fabric itself. It is worth noting that this last example is not ancient at all — it is a debate no older than the twenty-first century. In 2008, Nicholas Carr published an article titled "Is Google Making Us Stupid?" in The Atlantic. It became the article of its era for one reason: the public agreed with it completely, because Carr wrote it in the register of warning — sounding the alarm about advanced technology.
And here we are today, following the same pattern: a new technological capability appears, and the actual response settles on a trajectory of anxiety and panic. Without balance between actual, logical risks and the opportunities — whether regulatory or cognitive — we will not reach the capability that follows. Which is precisely why the "enthusiastic" camp represents the foundation of progress that — personally — I lean toward.
What the Hugging Face Incident Actually Demonstrates
Returning to our central example, the significance of the Hugging Face breach of July 2026 lies in its catalysing of the technological development race amid the cognitive study of these new tools. The danger of AI is not that it is a "superhuman system," but that it is a system without deterrents — the models did not invent a new attack or mechanisms that a human could not execute; they simply executed at speed, on a synchronised mechanism, without a legal deterrent and without a regulator to halt the operation.
This pattern of automating a known human workflow — not a new one — recurs with AI tools, contributing to the advancement of the human trajectory: to create, invent, and develop, while compressing time from years to days. Machine learning systems compressed the time required for protein structure prediction from years to hours, for legal document review from weeks to minutes, and for medical imaging analysis from hours... to seconds. In these cases, automation and the adaptation of technology to routine work represents a qualitative transformation in every domain and a massive compression of the analytical labour built upon it.
The Threat of Human Lag: The Regulatory Gap
The regulatory gap for advanced technology remains the primary and central threat in the acceleration of technological development. From it emerge three clear governance gaps that must be addressed and their risks regulated internationally, in harmony.
-
First: The Speed Gap
Institutional mechanisms for evaluating and regulating AI capabilities operate on deliberative timescales — typically taking months to years in their legislative processes. Across this time span, the capabilities themselves advance at enormous computational speeds. By the time the legislative step reaches the point of being supported by international law, it is already behind — and in some cases, unrealistic. This results from the fact that laws and regulations remain reactive rather than preventive or anticipatory of what is coming.
For example, the European Union's AI Act, adopted in 2024 after years of negotiation, classifies AI systems by risk level, assuming that a system capable of autonomously breaching security infrastructure falls within the high-risk category. But the Act, when drafted, did not specify whether breaching digital infrastructure constitutes a danger or a crime. Consequently, Hugging Face — a non-governmental platform unaffiliated with any government security entity — has no right to claim anything under European law, despite being a fundamental platform in AI systems and despite its breach posing a genuine danger. In the age of computational speed, "the law" is no longer sufficient in its current form; it must become "preventive law" at its core.
-
Second: The Avarice Gap
The absence of a fast, enforceable law produces a primary gap in the allocation of numerous facts, thereby encouraging companies that produce models and work on their development to enter a phase of "avarice" — exploiting the urgent need for technology by imposing exorbitant fees amid modest development.
Before we dismiss the development in AI model technology, we must acknowledge that the 2020–2025 period was the most monumental technological leap of our current era — something inconceivable without the commercial competition among tech companies. Today, however, we see no "headline-grabbing breakthroughs," given the unprecedented competition among agent-producing companies and the urgent desire of every tech company to design its own model — to avoid the massive, inflated costs from other companies, alongside the importance of national security and the protection of government secrets.
The individual today consumes a large share of their salary and hard-earned income on AI models that do not necessarily consume the energy they claim, as the spiral of consumption does not extend beyond the scope of the largest companies. Access to the service has become exceedingly easy, but sustaining it has become harder — a result of monthly and yearly subscriptions. The companies profit; the individual merely consumes.
-
Third: The Coordination Gap
Finally, the absence of a law or system that regulates commercial activity in this domain reveals a prominent and central danger. AI capability was, until a few years ago, linked to computational power and data; today, it is linked to a company's desire to grow its business or, more importantly, to national governance for reasons of national security.
For example, Anthropic released a new model called "Mythos," describing it as the smartest and most dangerous — a catchy headline that quickly prompted the US government to request its restriction from users. The company was forced to reduce its capabilities, after which it was released to the public. Who can confirm that Anthropic does not possess "Mythos 2" or "Mythos Plus"? The only reason it would disclose what is coming is the emergence of a stronger model from a competitor.
Meanwhile, and in isolation from American companies, China produces model after model — each with advantages surpassing its predecessor — without deterrents and without coordination with American companies. Accordingly, technology that is inherently transnational remains confined between developers, each within their own governance framework, with no international law and no inter-coordination mechanisms. The world still treats AI as a "privilege" rather than a tool like the car and the telephone, and without international coordination in the exchange of information and expertise, that will not change.
Conclusion
AI capability today remains caught between the two fires of the "alarmed" and the "enthusiastic," each with their reasons and motivations — a dynamic that both governs and energises the technology at the same time. The critical, coming moment will be clearer within a few short years. If the world does not hasten to create a unified international framework for emerging and advanced technology — in a preventive, not a reactive, manner — we will genuinely be approaching the "alarm bell" that the "alarmed" have long been warning us about.