The Sovereign Mind and the Machine - What the Legal Protection of the Forum Internum Owes to a Philosophy That Rejects It

Introduction: a paradox and its theoretical interest

Few philosophical works enjoy a posterity as asymmetrical as that of Ayn Rand. Marginal in philosophy departments, where she is rarely taught, she nonetheless occupies a considerable position within North American entrepreneurial culture. This influence has been documented by a now substantial body of literature. As early as 1996, Richard Barbrook and Andy Cameron identified in what they termed the “Californian ideology” an unstable amalgam between the counter-cultural heritage of the left and an antistatism of Randian inspiration1. Twenty-five years later, Adrian Daub argues that Rand’s heroic individualism has become an inseparable part of the way the technology industry presents itself to itself and to the world2.

What circulates in this manner is not a doctrine — objectivism as a system is rarely defended in its own terms — but a grammar: a set of figures and oppositions available for characterising a situation. The creator who produces stands opposed to the bureaucrat who extracts; value arises from the individual and not from the collective; the public norm is almost never presented as a technical error liable to be corrected, but as a penalty inflicted upon productivity. This grammar structures a significant part of contemporary discourse opposing the regulation of technology, and European regulation in particular.

Yet this same industry today deploys systems whose practical function, avowed in its own sales pitches, is precisely to spare the user the effort of judgment. If independence of judgment does indeed constitute the virtue that Rand saw in it, then generative artificial intelligence, in its dominant use, produces on a large scale exactly what The Fountainhead describes as the second-hand man.

The interest of this observation is not polemical. Pointing out an inconsistency between a discourse and a business model is a low-cost operation of little analytical yield. The interest lies elsewhere: the paradox brings to light an internal gap in the Randian position, and this gap is precisely the one that the law of fundamental rights occupies. The thesis defended here is the following. Objectivism posits the sovereign subject as a natural given, prior to any institution, which need only be left unhindered. This premise is empirically fragile. If, on the contrary, the independent mind is an achievement — presupposing determinate material, educational and environmental conditions — then the legal protection of the forum internum is not the limit of cognitive sovereignty, but one of its conditions of possibility.

The argument unfolds in four stages. It is a matter first of restoring the Randian position in its own terms (I), then of characterising precisely the mechanism by which generative systems affect the formation of judgment (II), before examining the state of the legal protection of the forum internum and its translation into the artificial intelligence regulation (III). The final section examines the limits of this protection and the actual scope of the proposed reversal (IV).

I. Volitional consciousness and the second-hand man

The position must be restored seriously, failing which the proposed reversal would be no more than a rhetorical device.

Rand’s ethical architecture rests on an explicit metaethical premise: life constitutes the sole ultimate standard of value, and reason is the fundamental means of survival proper to the human being. Reason is not, in this framework, one faculty among others; it is the organ by which the individual orients himself in reality and without which he could not subsist as a human being. From this follows the structure of the objectivist virtues, articulated around three cardinal values — reason, purpose and self-esteem — to which correspond three cardinal virtues: rationality, productiveness and pride3.

Independence figures among the virtues that proceed immediately from rationality. This derivation is not incidental, for it rests on a second premise, decisive for our purpose: consciousness is volitional. Thinking is an act that the individual must choose to perform and that he can, at every moment, refuse to perform4. The rational faculty does not function automatically; it presupposes an act of focusing that nothing guarantees. It is this thesis that gives the whole its particular tenor: the principal threat here is not error, but abdication.

The figure of the second-hand man, constructed in The Fountainhead around the character of Peter Keating, condenses this thesis5. Keating is not an individual who thinks badly; he is an individual who has substituted the reflection of others’ judgment for judgment of his own. He does not ask himself what is true, but what will be received as such. His downfall is not of an intellectual order but of a moral one, and it is voluntary.

This point deserves emphasis, because it is systematically omitted in the political uses of Rand. In the internal economy of the system, the primary danger is not external constraint. It is the abdication of judgment, of which external constraint is only a consequence: coercion prospers only on minds that have already renounced judging. The reading that makes objectivism a mere anti-statist doctrine thereby misses what constitutes its mainspring — a demand addressed to the individual before it is addressed to the State.

II. Anteriority: the mechanism proper to generative systems

The risk that generative artificial intelligence poses to the formation of judgment is not one of censorship. No content is forbidden, no position is prevented, and warnings framed in terms of the control of information miss the specificity of the phenomenon.

The mechanism is that of anteriority. The system formulates before the user has formulated. The uncomfortable moment in which thought must constitute itself without support — the cognitive equivalent of the blank page — is occupied by a text that is already articulated, plausible and syntactically coherent. All that remains is to approve or to correct. The cognitive cost of acceptance becomes marginal; that of disagreement presupposes reformulating, that is to say, performing the very work that the tool promised to spare. This imbalance is structural and not accidental: it follows from the very form of the interaction.

This displacement in the economy of intellectual effort is beginning to be documented empirically, with the precautions that a recent field calls for.

A study conducted by Microsoft Research and Carnegie Mellon University with 319 knowledge workers, covering 936 real professional use cases, establishes two convergent findings. On the one hand, the majority of respondents report reduced cognitive effort across the full range of activities associated with critical thinking — knowledge, comprehension, application, analysis, synthesis and evaluation. On the other, and this is the most significant finding, confidence placed in the system negatively predicts the exercise of critical thinking, whereas confidence in one’s own competencies predicts it positively6. In other words, the exercise of judgment decreases as confidence in the tool increases — the exact inverse of what most internal AI-governance policies presuppose, resting as they do on the hypothesis of systematic human verification.

A second study, conducted at the MIT Media Lab, adds a neurophysiological dimension. Fifty-four participants were divided into three groups — assistance by a language model, search engine, no tool — for a writing task, with measurement of brain connectivity by electroencephalography. The participants without a tool exhibited the most extensive networks, the search-engine users an intermediate engagement, and the language-model users the weakest connectivity. The latter also reported the lowest sense of ownership over their own texts and experienced greater difficulty in quoting their content7. These results must be handled with caution: the sample is small, the electroencephalographic methodology has been the subject of criticism, and the scope of the conclusions beyond the experimental setup remains disputed. They nonetheless converge with the literature on cognitive offloading.

Two clarifications are necessary in order to avoid misunderstandings.

First, the thesis is not that the machine “thinks in our place”. This convenient formula is false: it attributes to the system an operation of which it is devoid and shifts the question onto a metaphysical terrain of no interest. What is at issue is a more modest and more troubling phenomenon: it becomes less costly to validate than to elaborate.
Second, the pertinent distinction does not set tools against faculties. The calculator, the library catalogue or the search engine extend judgment without taking its place; they furnish materials to an operation that the individual retains. What changes here is that a system intervenes at the precise moment when judgment is forming and proposes an outcome before the question has been fully posed. This boundary is not technical — it cannot be read in the architecture of the system — but pertains to the manner in which the tool inserts itself into a practice. This is one of the reasons why the law has such difficulty grasping it.

III. The forum internum: from a theoretical right to a positive right

A. An absolute protection long without application

International law has protected freedom of thought since 1948, without ever having truly had to apply it.

Article 18 of the Universal Declaration, Article 18, paragraph 1, of the International Covenant on Civil and Political Rights, Article 9 of the European Convention on Human Rights and Article 10 of the Charter of Fundamental Rights of the European Union protect freedom of thought, conscience and religion. The structure of these provisions is constant and decisive: they distinguish the forum internum — thought itself — from its outward manifestation. Only the latter is susceptible to restriction. The Human Rights Committee explicitly confirmed this in its General Comment No. 22: the protection of the forum internum is absolute and admits of no limitation whatsoever, in any circumstances8.

This protection has nonetheless remained largely theoretical. Jan Christoph Bublitz has been able to describe freedom of thought as a “forgotten right”, and to suggest that it is perhaps the only human right devoid of practical application9. The reason for this desuetude is structural: the forum internum was inaccessible. No technology allowed it to be penetrated or directly altered, so that no litigation had occasion to specify its contours. A right whose violation is materially impossible generates no case law.

This premise no longer holds, and legal scholarship has seized upon the matter. Susie Alegre proposed as early as 2017 to rethink freedom of thought for the twenty-first century, before developing the argument in Freedom to Think10. Simon McCarthy-Jones has analysed the conditions of mental autonomy in the age of algorithmic systems11; Nita Farahany has documented the stakes on the side of neurotechnologies12; Bublitz and Reinhard Merkel have proposed a theory of harms to the mind and a right to mental self-determination13. The thematic report of the United Nations Special Rapporteur on freedom of religion or belief, Ahmed Shaheed, presented in October 2021, constitutes the first attempt at a UN systematisation of the content of this right. It draws out four attributes: the freedom not to disclose one’s thoughts, the freedom not to be penalised for one’s thoughts, freedom from any impermissible alteration of one’s thoughts, and the existence of an environment conducive to freedom of thought14.

This fourth attribute merits attention. It breaks with the purely defensive conception of the forum internum: freedom of thought is here no longer merely a space not to be violated, but a state to be made possible. We shall return to this.

B. Article 5 of the artificial intelligence regulation

It is in this context that the first positive translation of this protection at the scale of an internal market must be read.

Article 5 of Regulation (EU) 2024/1689, applicable since 2 February 2025, prohibits the placing on the market, the putting into service and the use of artificial intelligence systems employing subliminal techniques operating beyond the threshold of consciousness, or purposefully manipulative or deceptive techniques, having the object or the effect of materially distorting the behaviour of a person or a group by appreciably impairing their capacity to make an informed decision, in a manner that causes significant harm. Point (b) of the same provision targets systems that exploit vulnerabilities arising from age, disability or a specific social or economic situation15.

The Commission’s guidelines on prohibited practices, approved on 4 February 2025, specify that these prohibitions target practices contrary to the fundamental rights protected by Union law — foremost among them autonomy, human dignity and the integrity of the person — and recommend a case-by-case assessment attentive to context16. These guidelines are devoid of binding force, the authoritative interpretation of the regulation resting with the Court of Justice alone.

The legal characterisation of this provision deserves to be specified. Article 5 is not formally founded on Article 10 of the Charter, and the regulation falls under the internal-market legal basis. It nonetheless remains that the protected subject matter — the capacity to form a judgment without substantial alteration by a process operating outside the conscious field — coincides very precisely with the object of the forum internum. What the law of fundamental rights protected in an absolute manner but without an instrument, the law of the internal market now protects in a conditional but enforceable manner.

IV. The reversal and its limits

A. The infrastructure of cognitive sovereignty

Read from within the Randian grammar, the prohibition laid down by Article 5 constitutes a characterised infringement. It restricts the freedom to conduct a business and freedom of contract in the name of a diffuse harm, consented to by no one in particular, and entrusts to a public authority the task of assessing what, within an interface, pertains to legitimate persuasion or to prohibited manipulation.

Read from Rand’s own premises, it nonetheless protects exactly what she had erected into the condition of moral dignity.

This divergence signals a question that objectivism never poses head-on: what makes possible the sovereign subject that it posits? Rand treats independence of judgment as a natural disposition that need only be left unhindered — state coercion being the paradigmatic hindrance. This conception is not absurd, and it possesses an internal answer that it would be dishonest to ignore: for objectivism, the prohibition of the initiation of force and the guarantee of the right of property are precisely the institutional conditions of the exercise of reason. The system is therefore not devoid of a theory of conditions; it possesses one, but a deliberately minimal one, which retains as a harm to the mind only physical constraint or characterised fraud.

It is this reduction that poses a difficulty. It supposes that everything which is neither force nor fraud leaves judgment intact — a hypothesis that the work gathered in Section II makes difficult to sustain. The alteration of judgment can result from devices that exert no constraint, deceive as to no verifiable fact, and are adopted voluntarily. The fourth attribute drawn out by the Special Rapporteur — the requirement of an environment conducive to freedom of thought — names precisely the blind spot of this construction.

Hence the hypothesis that can be formulated: the independent mind is not a given but an institutional achievement. It presupposes an education that exercises it, available time that permits it, and an informational environment in which deliberation is not systematically pre-empted by actors with an interest in its outcome. The prohibition of manipulation does not delimit cognitive sovereignty; it constitutes one of its conditions of exercise. The regulator that the Randian grammar designates as the adversary of the sovereign creator works, on this precise point, to render the latter possible.

B. What the law does not grasp

This reversal must not, however, be presented as more complete than it is, and to omit this would be the argument’s principal defect.

The threshold laid down by Article 5 is high, deliberately so. The prohibition presupposes the conjunction of several cumulative conditions: a subliminal, manipulative or deceptive technique; a material distortion of behaviour; an appreciable impairment of the capacity for informed decision; a significant harm, actual or reasonably foreseeable. This threshold is justified — a lower standard would lead to characterising a considerable share of ordinary commercial communication as manipulation, and would pose considerable difficulties with regard to the principle of legal certainty.

It follows, however, that the phenomenon described in Section II largely escapes the scope of the prohibition. The erosion of judgment through the everyday, voluntary and non-deceptive use of convenient tools proceeds from no manipulative technique; it aims to cause no harm; it results from a service rendered, and rendered efficiently. The law grasps manipulation; it does not grasp convenience. And convenience is, in all likelihood, at the scale of actual practices, the most determining factor.

This gap circumscribes the scope of the thesis. The law cannot produce the independent mind; it can at most protect its external conditions against the most characterised infringements. On this point, the Randian position retains a measure of accuracy: no normative provision will ever dispense the individual from the act by which he chooses to judge. Regulation is a necessary condition, not a sufficient one — and it may be noted that the most pertinent instruments in this regard perhaps pertain not to prohibition, but to the AI-literacy obligation laid down by Article 4 of the regulation, whose scope remains largely to be constructed.

Conclusion

The question, then, is not whether machines will think in our place — they have neither the capacity nor the need. It is whether we will continue to choose to think once doing so has become optional.

On this precise point, Rand remains useful, and there would be some bad faith in not acknowledging it. If thought is a volitional act, it is always refusable, and no abundance of instruments will dispense with this choice. This is the serious part of her work. It turns, however, quite exactly against those who invoke it: for if independence of judgment is indeed the virtue she described, then protecting it against what corrodes it is not a hindrance to the sovereignty of the creator. It is what renders it possible.

Bibliography

Normative and institutional sources

Human Rights Committee, General Comment No. 22 on Article 18 of the ICCPR, CCPR/C/21/Rev.1/Add.4, 30 July 1993.
European Commission, Guidelines on prohibited artificial intelligence practices established by Regulation (EU) 2024/1689, C(2025) 884 final, Brussels, 4 February 2025.
European Convention for the Protection of Human Rights and Fundamental Freedoms, Art. 9.
Charter of Fundamental Rights of the European Union, Art. 10.
International Covenant on Civil and Political Rights, Art. 18.
Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act), OJ L, 2024/1689, 12 July 2024.
SHAHEED A., Interim report of the Special Rapporteur on freedom of religion or belief: freedom of thought, A/76/380, 5 October 2021.

Legal scholarship

ALEGRE S., “Rethinking Freedom of Thought for the 21st Century”, European Human Rights Law Review, No. 3, 2017, pp. 221-233.
ALEGRE S., “Regulating around Freedom in the ‘Forum Internum’”, ERA Forum, vol. 21, No. 4, 2021, pp. 591-604.
ALEGRE S., Freedom to Think: The Long Struggle to Liberate Our Minds, London, Atlantic Books, 2022.
ALEGRE S. and SHULL A., Freedom of Thought: Reviving and Protecting a Forgotten Human Right, Waterloo, Centre for International Governance Innovation, special report, 2024.
BUBLITZ J. C., “My Mind Is Mine!? Cognitive Liberty as a Legal Concept”, in E. HILDT and A. FRANKE (eds.), Cognitive Enhancement: An Interdisciplinary Perspective, Dordrecht, Springer, 2013, pp. 233-264.
BUBLITZ J. C., “Freedom of Thought in the Age of Neuroscience: A Plea and a Proposal for the Renaissance of a Forgotten Fundamental Right”, Archiv für Rechts- und Sozialphilosophie, vol. 100, No. 1, 2014, pp. 1-25.
BUBLITZ J. C. and MERKEL R., “Crimes Against Minds: On Mental Manipulations, Harms and a Human Right to Mental Self-Determination”, Criminal Law and Philosophy, vol. 8, No. 1, 2014, pp. 51-77.
FARAHANY N. A., “The Costs of Changing Our Minds”, Emory Law Journal, vol. 69, No. 1, 2019, pp. 75-110.
FARAHANY N. A., The Battle for Your Brain: Defending the Right to Think Freely in the Age of Neurotechnology, New York, St. Martin’s Press, 2023.
LIGTHART S., BUBLITZ J. C., DOUGLAS T., FORSBERG L. and MEYNEN G., “Rethinking the Right to Freedom of Thought: A Multidisciplinary Analysis”, Human Rights Law Review, vol. 22, No. 4, 2022, ngac028.
McCARTHY-JONES S., “The Autonomous Mind: The Right to Freedom of Thought in the Twenty-First Century”, Frontiers in Artificial Intelligence, vol. 2, 2019, art. 19 (corrigendum: Frontiers in Artificial Intelligence, vol. 4, 2021, 672279).
SENTENTIA W., “Neuroethical Considerations: Cognitive Liberty and Converging Technologies for Improving Human Cognition”, Annals of the New York Academy of Sciences, vol. 1013, No. 1, 2004, pp. 221-228.

Cognitive science and empirical studies

KOSMYNA N., HAUPTMANN E., YUAN Y. T., SITU J., LIAO X.-H., BERESNITZKY A. V., BRAUNSTEIN I. and MAES P., “Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task”, arXiv preprint arXiv:2506.08872, 2025 (version 2, 31 December 2025).
LEE H.-P., SARKAR A., TANKELEVITCH L., DROSOS I., RINTEL S., BANKS R. and WILSON N., “The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers”, in Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems, New York, ACM, 2025.

Philosophy and sociology of technology

BARBROOK R. and CAMERON A., “The Californian Ideology”, Science as Culture, vol. 6, No. 1, 1996, pp. 44-72.
DAUB A., What Tech Calls Thinking: An Inquiry into the Intellectual Bedrock of Silicon Valley, New York, FSG Originals x Logic, 2020.
MOROZOV E., To Save Everything, Click Here: The Folly of Technological Solutionism, New York, PublicAffairs, 2013.
ZUBOFF S., The Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power, New York, PublicAffairs, 2019.

Primary sources (objectivism)

RAND A., The Fountainhead, Indianapolis, Bobbs-Merrill, 1943.
RAND A., Atlas Shrugged, New York, Random House, 1957.
RAND A., For the New Intellectual: The Philosophy of Ayn Rand, New York, Random House, 1961.
RAND A., The Virtue of Selfishness: A New Concept of Egoism, New York, New American Library, 1964.