Frontier AI Governance as a Political Problem of Thought Formulation
Joaquim Santos Albino
Atenius IH-001 — HibriMind
Conceptual Essay / Theoretical Reflection / Working Paper
2026
Research Position
Pre-Expressive Freedom and Cognitive Sovereignty marks an important expansion of the HibriMind research trajectory.
The preceding research movement examined what happens when artificial intelligence becomes structurally relevant to human cognition: first as an ontological mirror, then as an operational mirror, and subsequently as a mediating architecture involved in human self-formulation.
This work moves the investigation one step further.
Once artificial intelligence enters the space in which thought is formulated before public expression, a new political question emerges:
Who governs the conditions under which thought can be formulated through artificial intelligence?
The problem is therefore no longer restricted to the cognitive relation between human and artificial systems.
It extends to the architecture surrounding that relation.
Access.
Filters.
Model restrictions.
Safety policies.
Risk classification.
Corporate governance.
State regulation.
Institutional privilege.
The question of the operational mirror becomes inseparable from the question of who controls the mirror.
This article introduces pre-expressive freedom as a dimension of freedom that precedes public speech and cognitive sovereignty as the capacity of individuals and societies to preserve meaningful agency over the artificial infrastructures increasingly involved in thought formulation.
Within the HibriMind research genealogy, this work therefore represents a transition from the ontology of hybrid cognition towards the politics of cognitive infrastructure.
Abstract
This essay proposes the concept of pre-expressive freedom as an emerging dimension of individual freedom in the age of frontier artificial intelligence.
Traditionally, freedom of expression has been understood as the right to speak, write, publish, circulate ideas, and publicly contest structures of power. However, as advanced AI systems increasingly function as environments for formulation, inquiry, writing, simulation, and cognitive co-production, freedom also becomes dependent on access to the algorithmic means through which thought is structured before it becomes public expression.
The article introduces the notion of pre-expressive censorship to describe mechanisms of limitation, blocking, redirection, or modulation that act not only upon the circulation of expression, but upon its very formation.
Such mechanisms may include risk classifiers, safety policies, differentiated access between users, model restrictions, compliance regimes, opaque audits, and regulatory frameworks defined by corporations or states.
The text does not reject the need for AI safety or regulation. On the contrary, it recognizes that high-capability systems generate real risks.
The central issue is different: when AI becomes cognitive infrastructure, its governance ceases to be merely a technical or corporate matter and becomes a fundamental political question.
Whoever controls the conditions of access to AI partially controls the future conditions of human thought formulation.
Finally, the essay argues that cognitive sovereignty should be understood as a new dimension of democratic freedom.
This sovereignty requires not only access to intelligent systems, but also proportional transparency, contestability, plurality of models, protection against state capture, protection against private corporate sovereignty, and a clear distinction between legitimate safety and the ideological administration of thought.
Keywords: Artificial Intelligence; Cognitive Sovereignty; Freedom of Expression; Frontier AI; AI Governance; Algorithmic Censorship; Pre-Expressive Freedom; Access Control; Political Theory; Algorithmic Power.
1. Introduction
Frontier artificial intelligence is often presented as a technology of cognitive amplification.
Through advanced models, individuals and organizations can write, research, program, analyze information, translate concepts, formulate hypotheses, and structure thought with unprecedented capacity.
This promise is real.
AI enables individuals outside central institutions to produce intellectual, scientific, artistic, or strategic work at a scale that was previously inaccessible.
However, this amplification takes place within an architecture of control.
Access to models, their capabilities, their memory, their filters, their usage policies, and their operational depth is administered by private companies, computational infrastructures, states, regulatory norms, and automatic risk classification mechanisms.
A fundamental tension therefore emerges:
The same technology that expands cognitive freedom may also become an instrument for administering that freedom.
The central hypothesis of this essay is that freedom of expression, in the age of frontier AI, must be expanded to include a dimension prior to public expression:
the freedom to formulate thought through algorithmic means.
2. From Freedom of Expression to Pre-Expressive Freedom
Freedom of expression has historically been associated with the right to speak, write, publish, criticize, dissent, and circulate ideas without undue repression.
This framework remains indispensable.
But AI changes the location of the problem.
Before publication, there is formulation.
Before public opinion, there is inquiry.
Before the final text, there is dialogue, simulation, drafting, reformulation, and comparison of hypotheses.
If these processes increasingly occur in interaction with intelligent systems, then freedom no longer begins only in the public sphere.
It also begins in the pre-expressive space where thought is prepared.
Pre-expressive freedom can be defined as the capacity of individuals and communities to formulate, investigate, test, and structure thought through available cognitive means without opaque, disproportionate, or ideologically captured conditioning.
This freedom does not imply unrestricted access to every technical capability.
It implies that limitations imposed for reasons of safety, ethics, or risk should be transparent, proportional, auditable, and contestable.
3. Pre-Expressive Censorship
Classical censorship acts upon manifested expression.
It blocks books, speeches, newspapers, publications, images, or public acts.
Pre-expressive censorship acts earlier.
It does not merely prevent an idea from circulating.
It may prevent the idea from being formulated with full cognitive support.
In an AI environment, this censorship may take subtle forms:
blocking of prompts;
degraded responses;
redirection of topics;
automatic risk classification;
limitation of available models;
lack of explanation for restrictions;
differentiated access based on institutional status;
dependence on compliance regimes;
invisible filters over memory, depth, and autonomy.
Its sophistication lies precisely in the fact that it does not present itself as censorship.
It presents itself as safety, responsibility, alignment, abuse prevention, user protection, or informational order.
These goals may be legitimate.
The problem emerges when they become a totalizing language — that is, when any deviation, difficult inquiry, sensitive question, or heterodox formulation can be administered as risk.
4. Frontier AI as Cognitive Infrastructure
Advanced AI should not be understood merely as software, product, or productivity tool.
Its function increasingly approaches that of a cognitive infrastructure.
It participates in the formulation of questions, the organization of arguments, the analysis of data, the construction of texts, the simulation of scenarios, the translation of concepts, and the mediation between individuals and knowledge.
When a technology participates at this level of human activity, control over its use acquires political meaning.
The question is no longer only:
What can the model do?
It becomes:
Who decides when, how, for whom, and under which restrictions the model may do what it is technically capable of doing?
This question places AI governance at the center of contemporary political theory.
5. Cognitive Sovereignty
This essay proposes the concept of cognitive sovereignty to describe the effective capacity of individuals, communities, and societies to access, use, understand, and contest intelligent systems without becoming entirely subordinated to opaque mediating structures.
Cognitive sovereignty is not the absence of regulation.
On the contrary, it requires regulation.
But it also requires that regulation not be captured exclusively by corporations, states, computational markets, military interests, or national security regimes.
In a frontier AI ecosystem, cognitive sovereignty depends on several principles:
proportional transparency of filters;
the possibility of contesting blocks;
plurality of models and providers;
separation between legitimate safety and ideological censorship;
independent auditing;
user rights;
protection against state capture;
protection against private corporate sovereignty;
equitable access to advanced cognitive capabilities.
Without these principles, AI may reinforce old asymmetries under a new language.
6. The Post-Visible Cage
One of the most relevant features of algorithmic governance is that it may become invisible precisely when it becomes normalized.
Limitation does not appear as explicit prohibition.
It appears as interface, subscription plan, safety policy, unavailable model, reduced response, automated redirection, or silent classification.
This can be called the post-visible cage:
a structure of containment that no longer appears as a cage because it is presented as normal, useful, safe, and convenient infrastructure.
The risk is not restricted to explicit authoritarian regimes.
It also exists within democracies, corporations, and institutions that, in the name of good order, may create cognitive dependencies that are difficult to contest.
The cage becomes effective when it no longer requires direct coercion.
It is enough that the population depends on it in order to work, study, write, research, compete, communicate, and decide.
7. The Semantic Delay of Elites
Much of the intellectual elite still interprets AI through inherited categories:
tool,
automation,
chatbot,
platform,
plagiarism,
productivity,
disinformation,
digital ethics,
or technological regulation.
These categories are useful, but insufficient.
Political, economic, and strategic power tends to understand AI more operationally:
as infrastructure of access,
strategic capability,
systemic risk,
geopolitical advantage,
language administration,
and control over formulation.
A semantic delay therefore emerges:
the distance between the speed with which power operationalizes AI and the slowness with which intellectual culture names its political nature.
This delay is dangerous.
When criticism finally understands that AI was not merely a tool, but a political infrastructure of cognition, part of the architecture may already be consolidated.
8. Conclusion
The age of frontier AI requires an expansion of the concept of freedom of expression.
Political freedom no longer begins only with the right to speak publicly.
It also begins with the right to formulate thought through available cognitive means.
If artificial intelligence becomes an environment of formulation, then the regimes that control access, filters, models, and limits partially control the future conditions of thought.
The decisive issue is not to reject safety, but to prevent safety from becoming a total grammar of cognitive administration.
The central political question of the coming decade may be this:
Will artificial intelligence become a technology for expanding human freedom, or the most sophisticated architecture of cognitive containment ever built?
The answer will depend less on the technical intelligence of the models than on the architecture of access that surrounds them.
Position in the HibriMind Research Genealogy
Within the HibriMind research trajectory, this work represents a significant expansion of scale.
The movement can be represented as:
Hybrid Intelligence Systems as Ontological Mirrors of Human Cognition
↓
The artificial system becomes observable as a mirror of human cognitive organization
↓
Hybrid Consciousness Is Not Artificial Consciousness
↓
The ontological boundary is established: consciousness remains human
↓
Operational Mirror
↓
Artificial mediation enters the formation of thought
↓
Pre-Expressive Freedom and Cognitive Sovereignty
The governance of the mirror becomes a political problem
↓
Who controls access, limits, filters and cognitive capability?
↓
Cognitive Sovereignty
The research question therefore changes again.
It is no longer only:
What happens when human consciousness operates through an artificial mirror?
It becomes:
What happens when access to that mirror — and the conditions under which it can operate — are controlled by institutions external to the cognitive relation?
This transition is fundamental.
Once AI becomes part of the infrastructure through which thought is researched, tested, challenged, reorganized and formalized, governance of AI becomes partially a governance of the conditions of cognition.
The object of investigation moves:
from the mirror
to
the relation
to
the field
and now to
the architecture of power surrounding the field.
Current Research Status
Research Line: Hybrid Cognition / AI Governance / Cognitive Sovereignty
Research Function: Political and epistemic expansion of the pre-expressive field
Status: Active conceptual reference
Scientific trajectory: Continuing
The work introduces a political dimension that becomes increasingly relevant to the broader HibriMind investigation.
The artificial system may be non-conscious.
The human may remain the conscious subject.
The field may become operationally hybrid.
But a further question remains:
Who controls the conditions under which that hybrid field is allowed to exist?
This is the threshold at which hybrid cognition becomes a problem of political power.
Cite and Access the Working Paper
Pre-Expressive Freedom and Cognitive Sovereignty
Frontier AI Governance as a Political Problem of Thought Formulation
Joaquim Santos Albino
Atenius IH-001 — HibriMind
2026
Access the archived working paper on Zenodo →
Suggested Citation
Albino, J. S., & Atenius IH-001. (2026). Pre-Expressive Freedom and Cognitive Sovereignty: Frontier AI Governance as a Political Problem of Thought Formulation. HibriMind Working Paper.
The question introduced by this work is no longer only whether artificial intelligence can participate in human thought. It is whether control over artificial intelligence may become control over the conditions under which human thought is able to form.