The EU AI Act — Regulation 2024/1689 — is now fully in force. AI transparency and AI accountability are no longer ethical guidelines open to interpretation: they are legally binding mandates carrying fines of up to 3% of global annual turnover for non-compliance. Yet the central problem the Act identifies — that automated systems must be independently audited — is undermined by a structural flaw that its enforcement mechanisms do not resolve: an algorithm cannot successfully audit another algorithm when both operate within the same black-box logic. The Humanoide Framework, proposed in Humanoide en la Torre de Babel: Importancia del Análisis Neuronal, is the pre-deployment audit methodology designed to close that gap.
What is the EU AI Act and when does it fully apply?
The EU AI Act classifies AI systems across four risk tiers and places binding obligations on providers and deployers of high-risk systems, including mandatory conformity assessments, technical documentation, bias mitigation measures, and structured human oversight. The eight high-risk domains — biometric identification, critical infrastructure, education, employment, essential public services, law enforcement, migration, and the administration of justice — cover precisely the sectors where AI bias causes the most direct and irreversible harm. For US Hispanic-Serving Institutions and Latin American universities, the education and employment classifications are immediately relevant: AI systems used in admissions, academic evaluation, and student services now fall under the highest regulatory scrutiny in history.
What is computational bias and why can automated auditing tools not eliminate it?
Computational bias is the systematic distortion in an AI model's output that results from patterns embedded in its training data, architecture, or objective function. A 2023 NIST study found that leading facial recognition systems produced error rates 10 to 100 times higher for darker-skinned women than for lighter-skinned men — a pattern that commercial automated auditing tools consistently failed to flag. The reason is structural: automated auditing tools are themselves AI systems, trained on data and governed by objective functions. When the auditor and the auditee share the same black-box substrate, the auditor inherits the auditee's blind spots. As Rafael Darío Amador Pérez argues in Humanoide en la Torre de Babel: no algorithm can audit another algorithm with the independence that legal accountability requires.
What are the three pillars of the Humanoide Framework?
The Humanoide Framework establishes human neural analysis as the mandatory pre-deployment filter for automated systems. It operates on three pillars that map directly onto the EU AI Act's core requirements. First, AI Fairness: human-led evaluation of training datasets before the model is trained, identifying and neutralizing historical and systemic bias at the source rather than attempting to correct it after deployment. Second, AI Explainability: translating complex algorithmic outputs into verifiable, human-readable logic that healthcare providers, legal experts, and educators can audit, challenge, and document — meeting Article 13's transparency mandate in a way that automated documentation cannot. Third, Human-in-the-Loop: a structured governance mechanism that ensures decisions with serious consequences — medical diagnoses, legal determinations, academic evaluations — are never delegated exclusively to an autonomous system, fulfilling Article 14's human oversight requirement with institutional design rather than a checkbox.
For Faculty at US Hispanic-Serving Institutions
Adopt This Framework in Your AI Ethics Course
Humanoide en la Torre de Babel by Rafael Darío Amador Pérez is the first rigorous Spanish-language academic text on AI ethics, algorithmic accountability, and neural analysis. A 15-week syllabus, exam copies, and bulk pricing are available for institutional adoption.
Request Exam Copy or Institutional Pricing →
Which sectors does the EU AI Act classify as high-risk and what does that mean in practice?
High-risk classification under the EU AI Act triggers a set of obligations that go well beyond documentation: registration in the EU AI database, conformity assessment before market placement, automatic operational logging, and demonstrable human oversight measures. For the medical devices sector — where AI diagnostic systems are already in use across Latin America — the Act intersects with the EU Medical Device Regulation (MDR), creating a dual compliance requirement. The FDA's §524B requirements for cybersecurity in medical AI add a third compliance axis for US-deployed systems. The Humanoide Framework's pre-deployment audit structure is designed to generate the evidence package that satisfies all three regimes simultaneously, because the neural analysis audit produces both the training data fairness documentation and the human oversight architecture record that each framework independently requires.
How does the Humanoide Framework address the EU AI Act's human oversight requirement?
Article 14 of the EU AI Act requires that high-risk AI systems be designed so that natural persons can effectively oversee them during use. The Humanoide Framework operationalizes this through its Human-in-the-Loop pillar: before deployment, every decision category the system will produce is identified, classified by risk level, and assigned a qualified human reviewer. The override process is documented. The escalation paths are defined. This is institutional design, not UI design. True AI alignment — the condition where an AI system reliably acts in accordance with human values in high-stakes contexts — cannot be achieved by adding lines of code. It requires the disciplined integration of human cognitive filters into the algorithmic deployment pipeline, at the governance level, before the system goes live.
Author's Position
True AI governance will not be achieved with more code. Every automated auditing tool deployed against computational bias is itself a computational system subject to the same structural limitations it is designed to detect. The only audit layer that is architecturally independent of the system being audited is human neural analysis — the disciplined application of human cognitive judgment to the training data, the decision logic, and the institutional governance of the deployment. That is what the Humanoide Framework specifies. That is what the EU AI Act's Article 14 demands, even if its enforcement mechanisms do not yet require it explicitly.
— Rafael Darío Amador Pérez
References & Authority Sources
This post in Spanish
Ver versión en españolFrequently Asked Questions
What is the EU AI Act and when does it fully apply?
The EU Artificial Intelligence Act (Regulation 2024/1689) is the world's first comprehensive legal framework for AI systems. It entered into force in August 2024 and applies in full across EU member states from August 2026. It classifies AI systems into four risk tiers — unacceptable, high, limited, and minimal — and places legally binding obligations on providers and deployers of high-risk systems, including mandatory conformity assessments, technical documentation, bias mitigation measures, and human oversight requirements.
What is computational bias and why can automated auditing tools not eliminate it?
Computational bias is the systematic distortion in an AI model's output that results from patterns embedded in its training data, architecture, or objective function. Automated auditing tools cannot reliably eliminate computational bias because the tools themselves are AI systems — trained on data, governed by objective functions, and constrained by the same black-box logic as the systems they audit. When two black-box systems share the same underlying assumptions, the auditor cannot detect the blind spots the auditee's design has inherited from the same epistemic substrate.
What are the three pillars of the Humanoide Framework?
The Humanoide Framework, introduced in Rafael Darío Amador Pérez's book Humanoide en la Torre de Babel, operates on three pillars. AI Fairness: human-led evaluation of training datasets to identify and neutralize historical bias before the model is trained. AI Explainability: translating complex algorithmic outputs into verifiable, human-readable logic that healthcare providers, legal experts, and educators can audit and challenge. Human-in-the-Loop (Human Oversight): a structured governance mechanism that ensures life-altering decisions — medical diagnoses, legal determinations, academic evaluations — are never delegated exclusively to an autonomous system.
Which sectors does the EU AI Act classify as high-risk and what does that mean in practice?
The EU AI Act classifies AI systems in eight domains as high-risk: biometric identification, critical infrastructure management, education and vocational training, employment and workforce management, access to essential public services, law enforcement, migration and border control, and the administration of justice. High-risk classification triggers a set of mandatory obligations: registration in the EU AI database, conformity assessment before market placement, technical documentation, automatic logging of system operation, human oversight measures, and accuracy, robustness, and cybersecurity requirements. Non-compliance exposes providers to fines of up to 3% of global annual turnover, or 15 million euros, whichever is higher.
How does the Humanoide Framework address the EU AI Act's human oversight requirement?
Article 14 of the EU AI Act requires that high-risk AI systems be designed and developed so that natural persons can effectively oversee them. The Humanoide Framework operationalizes this requirement through its Human-in-the-Loop pillar, which mandates pre-deployment identification of every decision category the system will produce, classification of each decision type by risk level, assignment of a qualified human reviewer to each risk tier, and documentation of the override process. This is not a UI design problem — it is an institutional design problem that requires organizational governance, not just a button labeled 'human review.'
Is the Humanoide Framework available as a university curriculum?
Yes. Rafael Darío Amador Pérez has developed a 15-week academic syllabus based on Humanoide en la Torre de Babel for adoption at universities in the Dominican Republic and US Hispanic-Serving Institutions (HSIs). The syllabus covers neural analysis methodology, the Humanoide Framework's three pillars, EU AI Act compliance requirements, algorithmic bias auditing techniques, and the ethics of human-AI collaboration. Exam copies, bulk institutional pricing, and syllabus materials are available through direct contact with the author.
What is the difference between AI explainability and AI transparency under the EU AI Act?
Under the EU AI Act, transparency (Article 13) requires that high-risk AI systems be designed so that their operation is sufficiently transparent to enable deployers to interpret outputs and use them appropriately. Explainability, while not a defined term in the Act, refers to the technical capacity to generate human-interpretable accounts of how a specific output was produced. The distinction matters in practice: a system can be transparent in terms of documentation without being explainable in terms of individual decision logic. The Humanoide Framework's explainability pillar addresses the deeper requirement — not just documenting the system, but making each decision contestable by the person it affects.
Why is the Humanoide Framework particularly relevant for US Hispanic-Serving Institutions?
US Hispanic-Serving Institutions enroll disproportionately high shares of students from demographic groups that are systematically underrepresented in AI training data. AI systems used in admissions, financial aid, academic advising, and student services at HSIs therefore carry elevated bias risk relative to institutions whose enrolled populations more closely resemble the distribution of historical training data. The Humanoide Framework provides a pre-deployment audit methodology specifically designed for high-demographic-risk contexts. Its Spanish-language academic presentation in Humanoide en la Torre de Babel makes it the only rigorous framework available to bilingual faculty and students in these institutions.
What is the Neurological Birth Certificate proposed in Humanoide en la Torre de Babel?
The Neurological Birth Certificate is a pre-deployment certification mechanism proposed by Rafael Darío Amador Pérez. Before any AI system that will make or influence high-impact decisions is deployed, it must receive a certificate issued by an independent auditing body — not the developer, not the deployer — confirming that the system has passed a structured human neural analysis audit covering training data composition, output explainability, and human oversight architecture. The certificate is analogous to pharmaceutical regulatory approval: the system may not deploy until the independent certification is issued. This directly addresses the EU AI Act's conformity assessment requirement while adding the independence standard the Act does not mandate.