Habeas Rogare
Inverting the Harm Principle and Cognitive Sovereignty in the Age of AI
1. The Crack in the Harm Principle and Habeas Rogare
In practice, Mill’s Harm Principle operates with an implicit presumption: the actor is considered legitimate until harm is proven, and the burden of proof falls on the affected party. For slow, cumulative, or invisible harms - where the user cannot observe their own cognitive erosion - proving individual damage becomes practically impossible.
Just as Habeas Corpus (1679) protects bodily integrity, Habeas Rogare names the fundamental right not to have one’s capacity for active questioning and reasoning eroded, without awareness or real choice, through the design of a third party.
Five independent empirical studies indicate the same direction: cognitive loss occurs when AI eliminates human cognitive participation (False Cognitive Power Transfer, FCPT), whereas pedagogically calibrated friction produces learning gains (CCPT).
2. Inverting the Presumption: Four Trigger Criteria
To avoid the trap of vague "social rights" that Mill himself rejected, the burden of proof shifts to the provider only when all four trigger criteria are satisfied:
- 1. Precondition of autonomy. The harm affects bodily integrity, core cognitive capacities (L2/L3 reasoning), or access to risk information - not mere offense, disagreement, or moral disapproval.
- 2. Plausible mechanism and empirical evidence. Supported by an explainable causal mechanism (effort aversion plus removal of epistemic friction) and convergent empirical findings.
- 3. Information asymmetry or lack of real choice. Users overestimate their performance with AI and cannot detect their own skill loss, while the friction alternative is absent or hidden in secondary menus.
- 4. Proportionality. The least restrictive effective measure applies, in strict order: risk disclosure and labeling, interface design obligations, and only as a last resort, prohibition.
3. Verification Across Precedents
Inverting the burden of proof is not unprecedented; it already governs harms that are diffuse, cumulative, and invisible to the recipient:
| Criterion | Leaded Petrol | Secondhand Smoke | Zero-Friction AI |
|---|---|---|---|
| 1. Autonomy precondition | Cognitive capacity | Bodily integrity | Reasoning (L2, L3) & self-assessment |
| 2. Mechanism & evidence | Documented neurotoxicity | 1986 Surgeon General report | Effort aversion + 5 convergent studies |
| 3. Asymmetry / no choice | Producers knew; air unavoidable | Known risk, unavoidable nearby | Loss invisible to user; friction hidden |
| 4. Proportional measure | Phase-out / ban | Warnings & space restrictions | Visible Cognitive Slider & justified default |
4. From Request to Legal Obligation: Habeas Rogare Act
Modeled on the European REACH principle (“no data, no market”), the inversion turns an ethical plea into an alternative legal duty:
- Article 1 (Alternative Obligation). Providers of general-purpose generative AI must either (a) prove through independent studies that their zero-friction default does not reduce users’ unassisted reasoning capacity, or (b) implement a visible, gradual cognitive friction control (MCS / Cognitive Slider).
- Article 2 (Main-Screen Visibility). The friction control must sit on the main session screen - not hidden in settings or dependent on prompt-engineering skill.
- Article 3 (Minors & Education). For minor accounts and educational deployments, the default setting must be an active friction level (calibrated by teachers under Labora. Roga. Proba.), never OFF.
- Article 4 (Risk Disclosure). Selecting zero-friction mode displays a clear notice warning that prolonged use without personal cognitive effort may reduce unassisted problem-solving capacity.
- Article 5 (Liability & Enforcement). Links non-compliance to market surveillance fines, misleading omission rules (Directive 2005/29/EC), and product defect presumptions (Directive (EU) 2024/2853).
What this proposal does NOT do: it does not ban zero-friction AI, nor does it impose a mandatory friction level on informed adults. Position OFF remains freely available.
Full Paper & Citation
Stan, A. (2026). Habeas Rogare: Reversing the Perspective in the Harm Principle and Cognitive Sovereignty in the Age of AI (v2.3). Zenodo. DOI: 10.5281/zenodo.24105279