2050 Reader Resources

Tools for the Reader

These pages do not extend the narrative: they make its method available for reference. The map distinguishes among the three controlled vocabularies; the worksheet preserves evidence and decisions throughout the life cycle; the advanced protocol addresses two-system checks. Next come Davide’s example, the note on the characters, the glossary, the timeline, and the maps of the book’s themes.

Where necessary, the entries distinguish among three layers: documented historical and technical facts, narrative scenarios set in 2050, and the book’s original proposals. The author also used AI tools to cross-check sources and references; source selection and responsibility for every claim remain with the author.

Map of the Three Vocabularies

The three lists answer different questions. The mode describes the power exercised by the system; the evidence status qualifies each question in the Test; the organizational decision determines what is authorized overall. They are not added together, and one does not automatically convert into another.

From the Test to the Decision

Assists does not mean Proceed; Sufficient does not mean a favorable judgment. If the model, data, provider, scale, or responsible people change, the Test must be reopened.

Fillable Worksheet for the 2050 Test

Six pages: identification; seven questions; monitoring; decision and review. Complete with date-stamped facts. Sufficient applies only within the written scope; Unresolved requires evidence and a deadline; Blocking narrows or stops authorization. Benefits do not compensate for the absence of Redress, Exit, or authority to stop.

The Seven Questions—Purpose and Mandate

The Seven Questions—Evidence and Distribution

The Seven Questions—Harm and Redress

The Seven Questions—Exit and Monitoring

Transition to Life-Cycle Monitoring

Decision and Conditions

Advanced Two-System Testing Protocol

A second system can look for errors or disconfirming conditions; it does not constitute independent verification merely because it has a different name, model, or vendor. It is supporting material for the Evidence—not a second Test, and not a way to add votes.

Before use, define the legitimate outcome, prohibited actions, exposed groups, the time-to-harm window, the divergence that can change the outcome, the responsible party, and the stop condition. Test in shadow mode and with adversarial cases; only afterward, if justified, test at limited volume with reversible effects. The second system does not by itself resolve conflicts of values, disparities in access, insufficient organizational capacity, or the absence of Redress.

Completed Example: Davide, October 15, 2050

The minutes distinguish the evidence status from the organizational decision. Even with all seven questions documented, a capable product may receive Do not delegate when the remaining benefit does not justify the oversight, dependence, and exit costs in the specific context.

Reopening: April 15, 2051

Six months later, the no is reviewed, not erased. The proposal is new and narrower; the test remains separate from the purchase and preserves the concrete possibility of saying no again.

Note on the Characters

The scenes take place around 2050, over a span of several years, and are arranged by theme, not necessarily in chronological order. The characters are not real people: they are composite narrative figures created to show choices, consequences, and limits.

Principal Characters

Sofia —

A young adult and Amir’s older sister. She moves from university into her first consulting role and learns to distinguish the availability of answers from the quality of judgment.

Amir —

Sixteen years old, Sofia’s younger brother, and a student. Through generative music, he confronts how easy it is to produce and how hard it is to verify without giving up the desire to say something.

Davide —

An experienced professional who runs a small business. He takes a twin for the contracting process from testing to nonadoption, accepts the cost, and six months later reopens only a narrower proposal, without turning the first no into a final verdict.

Luca —

A hospital physician. He meets Teresa along her care pathway and uses tools capable of seeing patterns the human eye may miss, while preserving the evidence, time, and authority needed to correct or stop the clinical decision.

Elena —

The mayor of a midsize city. In logistics projects and on the digital campus, she separates observation, identification, and authorization, negotiates thresholds and stop provisions, and does not treat the loss caused by prudence as a victory.

Karim —

A logistics worker and resident of an urban neighborhood. He makes visible the effects of metrics and mobility systems on safety, income, and movement, while also bearing the temporary cost of Redress.

Teresa —

Seventy-eight years old, with heart disease and diabetes, and widowed for a little over a year. She lives alone with unobtrusive sensors, distinguishes assistance from presence, and preserves the authentic archive without activating the simulation of her husband.

Matteo —

A civic communications professional in a midsize city. He is neither a journalist nor a politician: he is accountable for the provenance, publication, and correction of the content his office signs off on.

Irene —

A city council member who supports the logistics hub and campus projects. She makes the present cost of waiting visible, demands verifiable timelines and consequences, and preserves a documented dissent that can be reopened.

Asha Said —

Operations manager at a partner company in Mombasa. In Chapter 14, she measures continuity and exit from the point where a digital order becomes a real delivery, with authority to suspend new pickups.

Characters in Interlude II

Marta —

The head of a twelve-person technology provider. She shows how the cost of control can favor those with greater administrative capacity and proposes a proportionate exit test without denying the competitive gap.

Lidia —

The public procurement lead in the same case. She seeks proportionate evidence without weakening comprehensibility, exit, or authority to stop, and keeps the case file operational even when oversight carries a cost.

Glossary

A guide to the technical terms used in the book, explained in terms of what they change in everyday life. The definitions do not replace standards, regulations, or specialist manuals: they clarify the sense in which these words are used in these pages.

Artificial intelligence (AI) —

A set of techniques that allow a system to perform tasks associated with human capabilities, such as recognizing, predicting, translating, recommending, or generating. A convincing performance does not by itself demonstrate consciousness, experience, or understanding.

Machine learning —

An approach in which a model modifies its parameters based on examples, rather than receiving only rules written one by one. What it learns depends on how the problem is framed, the data, and the criteria by which it is evaluated.

Deep learning and neural networks —

A neural network is a mathematical structure of connected units and parameters that can be modified during training. Deep learning uses multilayer networks to construct useful representations of images, speech, language, and other signals. The name evokes the brain, but it does not mean that a small brain is hidden inside the computer.

Transformer —

An architecture that became influential with the 2017 paper “Attention Is All You Need.” It uses self-attention mechanisms to weight the relationships among positions in a sequence and makes it possible to process many of them in parallel during training. It does not “read” a text as a person does.

Large language model (LLM) —

A neural model trained on large collections of text to estimate which tokens may follow the context currently available. During generation, it does not necessarily choose the most likely alternative every time: instructions, context, the selection method, and later stages of training all play a role. Fluency does not guarantee truth or understanding.

Token —

A conventional unit into which a model divides text: it may correspond to a word, part of a word, a punctuation mark, or another segment. In this book, the term is used only in its linguistic sense.

Training data and datasets —

A dataset is a collection organized for analysis, training, or evaluation; training data are the examples used to modify a model’s parameters. Coverage, provenance, labels, and omissions depend on choices about what to observe, include, and exclude, and they influence capabilities and errors.

Hallucination —

A plausible but false or unsupported response, produced with the same fluency as accurate information. This is a term of art, not an indication that the system is having a perceptual experience.

Algorithm —

A procedure made up of steps or rules for transforming data and obtaining an outcome. Even when formalized, it incorporates choices about objectives, categories, thresholds, and exceptions.

Algorithmic bias —

A systematic distortion that may arise from the data, labels, objective, or context of use and produce unequal effects among people or groups.

Profiling —

The processing of data to evaluate or predict aspects of a person, such as preferences, behavior, reliability, or risk. Its lawfulness and limits depend on the context and the applicable rules.

Predictive surveillance —

The use of data and models to anticipate behaviors or risks and guide monitoring or interventions. It can turn a prediction into present treatment of a person and requires limits, verification, and contestability.

Unmeasured interval —

A temporary space in which a person can hesitate, contradict themselves, or change without every variation immediately producing an operational profile. It is neither the complete absence of technology nor a guarantee of freedom: continuous collection requires a defined harm, a proportionate purpose, and a verifiable expiration date.

Distributed ledger or blockchain —

A distributed ledger preserves a state shared among multiple parties; a blockchain is one possible structure for such a ledger. It can help independent organizations coordinate without a single administrator, but it does not make entered data true, eliminate the need for trust, or constitute a requirement for digital identity, verifiable credentials, or selective disclosure.

Verifiable credential —

An attestation composed of relevant attributes, issued by an accountable issuer, and secured so that a verifier can check its provenance and integrity. Technical verification does not make a false fact true: the quality of the issuer, procedures, credential status, and the ability to correct or revoke it all matter.

Selective disclosure —

The presentation of only the necessary attributes contained in a credential, while keeping the others hidden. It reduces the excess disclosure of information, but it does not by itself guarantee anonymity or prevent different interactions from being linked.

Zero-knowledge proof —

A cryptographic proof that makes it possible to demonstrate possession of information or satisfaction of a condition without revealing the underlying data. It is useful in some contexts, but it does not replace sound privacy and accountability architecture.

Digital wallet —

A tool through which a person can store, present, and manage digital credentials or attributes. It must provide for security, accessibility, recovery, revocation, delegation, and assistance; giving the person control must not become abandonment.

Digital identity —

The set of means and attributes through which a person or organization is recognized in systems. It is not the whole person or a single account. Identification, authentication, and authorization answer different questions.

Authentication and authorization —

Authentication verifies factors associated with a claimed identity, such as a credential, a device, or several factors combined. Authorization establishes what that party or system may do in the context. Passing the first does not automatically confer the second.

Interoperability and portability —

Interoperability allows different systems to exchange data and recognize functions or attestations according to common rules. Portability keeps data, identity, permissions, configurations, and logs usable outside the primary environment; it is incomplete without recovery, revocation, assistance, and bounded delegation. Neither one by itself guarantees continuity.

Data minimization —

A principle of the GDPR under which personal data must be adequate, relevant, and limited to what is necessary in relation to the purposes for which they are processed.

Proof minimization —

A design proposal in this book: demonstrate only what a decision requires, without handing over the person’s entire biography. It extends the logic of proportionality to the design of digital verification. See also: Verifiable credential; Selective disclosure; The 2050 Test.

Cloud —

A set of remote computing resources used to host data and services. The word suggests lightness, but the cloud depends on buildings, networks, contracts, jurisdictions, expertise, and energy.

Data center —

A physical facility that houses servers, networks, cooling systems, and backup power. It is a material, territorial, and organizational part of digital infrastructure.

Application programming interface (API) —

An interface that establishes how two software systems can request data or actions from one another. APIs make integration possible and, at the same time, create technical and economic dependencies.

Internet of Things (IoT) —

A network of physical objects equipped with sensors, software, and connectivity, capable of collecting data or receiving instructions: medical devices, meters, vehicles, traffic lights, appliances, and many others.

Deepfake —

A synthetic image, audio recording, or video that convincingly imitates a real person. Provenance and transformations can aid verification, but no single technical signal by itself establishes the whole truth.

Epistemic fatigue —

A condition in which the cost of reconstructing a conclusion grows to the point that a person suspends a judgment that remains open to revision or withdraws from participation. It is not the same as credulity or disinterest: it also depends on interfaces, access to sources, time, and the ability of institutions to make verification practicable.

Digital twin —

A computational representation of an object, process, or system, updated with data to observe or simulate its possible states. It does not necessarily imply an agent capable of acting on behalf of a person.

Operational twin —

A proposal in this book: a software agent tied to a principal and to an observable Mandate. It operates in Assists, Proposes, Decides, or Executes mode only within declared tasks, data, duration, thresholds, and authority to stop. When it functions as a tutor twin, it makes steps, errors, and alternatives visible so that competence can grow and the need for assistance can diminish. Capability, reliability, and authority remain distinct.

The 2050 Test —

A non-compensatory decision method proposed in this book. It organizes, in this order, Purpose / Mandate / Evidence / Distribution / Harm / Redress / Exit. Each answer receives an evidence status—Sufficient / Unresolved / Blocking—and leads to an organizational decision—Proceed / Limit and test / Do not delegate / Stop. It neither assigns a score nor certifies a product: a Blocking condition is not offset by the other benefits.

Operational mandate —

A written scope that connects a principal to a mode—Assists / Proposes / Decides / Executes—and specifies tasks, data, people affected, locations, duration, volume, logs, thresholds, prohibited actions, and authority to stop. If the model, data, provider, scale, task, or responsible parties change, the Mandate must be narrowed or reviewed; it is not the same as technical capability.

Technological sovereignty, strategic dependence, and infrastructural voice —

Sovereignty is the relative capacity to understand, choose, negotiate, and replace dependencies, not autarky. A dependence becomes strategic when it diminishes a capacity before an alternative practical path exists. Infrastructural voice is the concrete power to name the harm, suspend, correct, export, and replace, not consultation without consequence.

Right to erasure, also known as the right to be forgotten —

Under the GDPR, the right to obtain the erasure of personal data when the prescribed conditions are met, subject to limits and exceptions. The GDPR does not apply to the personal data of deceased persons, and national rules differ. The post-mortem digital directive—express consent, scope, revocation during life, and shutdown—is a proposal in this book, not a uniform right already in force.

Authentic archive, memorial, post-mortem simulation, and impersonation —

The archive preserves original traces; the memorial organizes them without speaking as the deceased person; the generative simulation produces new responses or representations from the data; impersonation deceptively presents them as coming from the real person. Ownership of the files, kinship, or payment does not amount to a Mandate to generate a first-person voice.

Verified comprehensibility —

The ability of a person or independent body to reconstruct why a system produced an outcome, test the system on an ordinary case and an edge case, and use that verification before the consequence occurs. It is not the same as a generic explanation or a formally complete case file: if the verification cannot change or stop the outcome, the Mandate must be narrowed.

Structural opacity —

A form of opacity that emerges when many components, updates, agents, and organizations combine in such a way that no actor possesses the complete picture in time to control the outcome. Opening the code may help without resolving it. See also: Verified comprehensibility; Time to harm and control window; Oversight budget.

Capability, reliability, and authority —

Three different questions: what a system can do, how consistently it does so under real conditions, and what it is permitted to do on someone’s behalf. Capability and reliability do not by themselves create authority or require adoption or complete delegation.

Human oversight —

Oversight is meaningful only when an identifiable person has readable evidence, time, competence, and authority to modify or stop the action, and when the handoff, reasons, and subsequent review remain on record. The formal presence or signature of an operator is not enough.

Meaningful recourse —

The concrete ability to receive an understandable notice, identify and correct the relevant elements, and reach without obstacles an office capable of suspending the effect, reopening the case, and communicating a reasoned outcome within a timeframe compatible with the harm.

Residual harm and operational case file —

An honest application of the Test does not promise zero harm. The case file retains value when it distinguishes original sources, human annotations, and generated summaries, acknowledges what cannot be reconstructed, and connects evidence, remedial actions, and restrictions. Its purpose is to learn and correct, not to declare control perfect.

Time to harm and control window —

Time to harm separates the error from its consequence; the control window is the time actually available to recognize, reconstruct, and intervene. Oversight is meaningful only if the necessary operations fit within that window; otherwise, it becomes nominal.

Oversight budget —

The relationship between the volume and speed of the decisions requiring oversight and the human resources available to review them attentively. When the load persistently exceeds the capacity for verification, human intervention becomes a belated signature. See also: Human oversight; Verified comprehensibility; Degraded mode.

Continuity, degraded mode, and residual competence —

Continuity preserves an essential function while also recording the costs transferred to users and staff. Degraded mode provides a deliberately narrower service, with declared limits, during an outage. Residual competence is the local knowledge and authority needed to activate it, manage exceptions, and reconcile the work when the system returns.

Exit test —

A practical verification, before adoption and then periodically, that data, logs, configurations, and permissions can be exported, reconstructed, used in degraded mode, and restored to operation without the primary provider. A portability clause is not enough if the organization cannot resume work, record losses, and return without hidden dependence.

Content provenance —

Verifiable information about a piece of content’s acquisition, origin, transformations, and approvals. These records distribute responsibility and help reconstruct a technical history, but they do not by themselves certify sincerity, context, or truth; incomplete metadata do not automatically make a document false.

Living license —

A proposal in this book for the use of a voice, image, or work in generative systems: consent bounded by purpose and duration, traceability of uses, compensation where applicable, the possibility of revocation, and treatment of derivative works. It does not describe a uniform license already recognized under current law. See also: Content provenance; Operational mandate; The 2050 Test.

Life cycle assessment (LCA) —

A method for assessing environmental impacts across declared stages and boundaries, from raw materials to end of life. Results depend on the goal, functional unit, data, and scope and are not comparable if these elements change without disclosure.

Material metabolism —

A way of reading digital infrastructure through what enters, circulates, and remains: energy, water, minerals, equipment, labor, emissions, and waste. It is a lens proposed in this book, not a single metric; comparisons require declared boundaries, units, and locations. See also: Life cycle assessment; Water withdrawal and consumption; Technological sovereignty.

Water withdrawal and consumption —

Withdrawal is water removed from a source; consumption is the share not immediately returned to the local environment. Confusing them makes comparisons among sites, technologies, and seasons difficult to interpret.

Solely automated decision-making —

Within the meaning of Article 22 of the GDPR, a decision made without meaningful human intervention that produces legal effects concerning a person or similarly significantly affects that person. Its scope, exceptions, and safeguards depend on the applicable law.

High-risk AI system —

A legal category under the AI Act that applies to systems identified according to the function, context, and conditions established by the regulation. It is not a generic synonym for a dangerous system and carries specific obligations that take effect progressively.

General-purpose AI model (GPAI model) —

A model capable of performing a wide range of tasks and being integrated into different systems. Under the AI Act, it is a regulatory category distinct from the individual system that incorporates it.

Essential Timeline

This timeline selects the milestones needed to navigate the book and classifies each entry as a historical fact or synthesis, a regulatory fact, a narrative scenario, or an author’s proposal.

1950 —

Historical fact. Alan Turing publishes “Computing Machinery and Intelligence” and replaces the abstract question “Can machines think?” with the imitation game, shifting attention to observable behavior.

1955–1956 —

Historical fact. John McCarthy, Marvin Minsky, Nathaniel Rochester, and Claude Shannon prepare the proposal for the Dartmouth summer workshop; the term artificial intelligence becomes the banner of a new field of research.

Late 1950s —

Historical synthesis. Frank Rosenblatt’s perceptron fuels the idea that a machine can learn from examples, while hardware, data, and methods remain limited.

1966 —

Historical fact. Joseph Weizenbaum introduces ELIZA, a simple program capable of producing, through rules and keywords, a surprising impression of listening.

1969 —

Historical fact and synthesis. Marvin Minsky and Seymour Papert analyze the limitations of simple perceptrons. The AI winters that follow also stem partly from the gap among promises, funding, computing power, and results.

1969 —

Historical fact—Infrastructure. ARPANET transmits the first message between two nodes, making a packet-switched network a reality from which broader communication architectures will develop.

January 1, 1983 —

Historical fact—Infrastructure. ARPANET completes the transition to TCP/IP: different networks can communicate through a common family of protocols, a decisive step toward the Internet as a network of networks.

1986 —

Historical fact. David Rumelhart, Geoffrey Hinton, and Ronald Williams provide an influential demonstration of how error backpropagation can train multilayer networks.

1989–1991 —

Historical fact—Infrastructure. Tim Berners-Lee proposes the World Wide Web at CERN in 1989, develops its first browser and server, and in 1991 the project begins to move beyond the laboratory, opening the network to hypertext navigation.

1997 —

Historical fact. Deep Blue defeats Garry Kasparov in a six-game match, making the power of computation in a domain with closed rules visible to the general public.

From the 1990s into the early 2000s —

Historical synthesis. The Internet, the Web, and mobile services transform the network from a specialized environment into an everyday infrastructure; archives, identities, communications, and services begin to depend on interconnected systems.

2012 —

Historical fact. AlexNet achieves a markedly better result in the ImageNet competition using graphics processors, making the acceleration of deep learning in image recognition clear.

2017 —

Historical fact. The paper “Attention Is All You Need” introduces the Transformer architecture, which will become central to large language models and many generative systems.

November 30, 2022 —

Historical fact. ChatGPT is introduced to the public as a research preview and makes conversational AI accessible for everyday use.

The 2010s and 2020s —

Historical synthesis. Cloud, data centers, APIs, platforms, and connected devices combine into increasingly invisible infrastructures on which work, health care, government, mobility, and information depend.

August 1, 2024 —

Regulatory fact. Regulation (EU) 2024/1689, known as the AI Act, enters into force. Its provisions apply progressively: this date does not mean that all obligations take effect simultaneously.

Around 2050 —

Narrative scenario. The vignettes, organized by theme, span several years. The fourteen- and eighteen-month intervals belong to Asha’s storyline; the six-month reviews concern Marta/Lidia and Davide; future dates, costs, timeframes, quantities, capabilities, and geopolitical arrangements are scenario parameters, not predictions or observed values. The scenarios include projects that are limited, halted, and not adopted.

The Book’s Proposals —

Author’s proposal. The 2050 Test, proof minimization, the operational twin, the living license, and the post-mortem digital directive are tools to be tried, revised, or abandoned: they are not established facts, uniform law, or predictions.

Topic Index

The book returns to the same questions at different scales. This map shows where each theme is developed and uses stable titles, so the cross-references remain valid when page numbers change.

Index of Names, Laws, and Standards

A compact index of substantive references in the main text and back matter. It complements the Topic Index without reintroducing the comprehensive index from previous editions.

People

• Amir 31, 136, 149
• Berners-Lee, Tim 300
• Davide 57, 95, 260
• Elena 69, 202, 265
• Hinton, Geoffrey 300
• Irene 76, 201, 266
• Karim 60, 71, 77
• Kasparov, Garry 300
• Lidia 189, 193
• Luca 105, 113, 267
• Marta 189, 192
• Matteo 131, 139, 265
• McCarthy, John 299
• Minsky, Marvin 299
• Papert, Seymour 299
• Rochester, Nathaniel 299
• Rosenblatt, Frank 299
• Rumelhart, David 300
• Said, Asha 219, 221, 226
• Shannon, Claude 299
• Sofia 11, 83, 266
• Teresa 114, 239, 247
• Turing, Alan 12, 299
• Weizenbaum, Joseph 13, 299
• Williams, Ronald 300

Laws and Regulatory Frameworks

• AI Act—Regulation (EU) 2024/1689 181, 301, 311
• Copyright in the Digital Single Market—Directive (EU) 2019/790 313
• Digital Services Act—Regulation (EU) 2022/2065 313
• European Digital Identity—Regulation (EU) 2024/1183 309
• GDPR—Regulation (EU) 2016/679 292
• data of deceased persons and Italian law (Art. 2-terdecies) 244
• solely automated decision-making (Art. 22) 310
• In vitro diagnostic medical devices—Regulation (EU) 2017/746 311
• Medical devices—Regulation (EU) 2017/745 311
• Platform work—Directive (EU) 2024/2831 310

Standards, Specifications, and Protocols

• ISO 14040 and ISO 14044—life cycle assessment 315
• NIST SP 800-63-4—Digital Identity Guidelines 309
• OpenID for Verifiable Credential Issuance 1.0 (OID4VCI) 309
• OpenID for Verifiable Presentations 1.0 (OID4VP) 309
• TCP/IP 300
• W3C Verifiable Credential Data Integrity 1.0 309
• W3C Verifiable Credentials Data Model 2.0 309

Endnotes

These endnotes document the historical, technical, and regulatory points that require a source. The scenes set in 2050 and the author’s proposals are not presented as present-day facts. Each note appears under the relevant chapter; the references remain readable and are accompanied by the stable [SRC-…] code for the corresponding entry in the Documentary Bibliography.

Chapter 1

• 1. Turing, Dartmouth, and the terminology of artificial intelligence. The imitation game is described on the basis of A. M. Turing’s 1950 paper; the 1955 Dartmouth proposal documents its signatories, conjecture, and programmatic use of the term artificial intelligence. Principal references: Turing (1950) [SRC-TURING-1950]; Dartmouth proposal (1955) [SRC-DARTMOUTH-1955].
• 2. The perceptron and AI winters. The text distinguishes the theoretical limitations analyzed by Minsky and Papert from hardware, data, funding, and expectations. Principal references: Rosenblatt (1958) [SRC-ROSENBLATT-1958]; Minsky and Papert (1969; expanded edition 1987) [SRC-MINSKY-PAPERT-1969].
• 3. Backpropagation, Deep Blue, AlexNet, and the Transformer. These milestones are documented, respectively, by the work of Rumelhart, Hinton, and Williams; the IBM archive; the work on AlexNet; and “Attention Is All You Need.” Principal references: Rumelhart, Hinton, and Williams (1986) [SRC-RHW-1986]; IBM archive on Deep Blue (1997) [SRC-IBM-DEEPBLUE]; Krizhevsky, Sutskever, and Hinton (2012) [SRC-ALEXNET-2012]; Vaswani et al. (2017) [SRC-TRANSFORMER-2017].
• 4. ELIZA and ChatGPT. ELIZA’s rule-based operation is documented in Weizenbaum’s paper; the date November 30, 2022, and its designation as a research preview come from OpenAI’s official documentation. Principal references: Weizenbaum (1966) [SRC-ELIZA-1966]; OpenAI (2022) [SRC-OPENAI-CHATGPT].

Chapter 2

• 5. Verifiable credentials, signatures, and wallets. The issuer–holder–verifier roles, proof integrity, and presentation methods are documented in W3C recommendations and OpenID specifications. A signature verifies a key and integrity; binding the key to an entity requires a trust system. For the distinction between a distributed ledger and a blockchain, see also NIST’s technical overview. Principal references: W3C, Verifiable Credentials Data Model 2.0 (2025) [SRC-W3C-VC20]; W3C, Data Integrity 1.0 (2025) [SRC-W3C-DI]; W3C, JOSE and COSE for verifiable credentials (2025) [SRC-W3C-JOSE]; OpenID Foundation, OID4VCI 1.0 (2025) [SRC-OID4VCI]; OpenID Foundation, OID4VP 1.0 (2025) [SRC-OID4VP]; NISTIR 8202 (2018) [SRC-NIST-8202].
• 6. Data minimization, selective disclosure, and European digital identity. The GDPR requires data to be adequate, relevant, and limited to what is necessary; the European Digital Identity Regulation governs voluntary wallets and selective disclosure. The Commission’s implementation framework, updated June 22, 2026, retains the deadline for national wallets to be issued by the end of 2026; the official Architecture and Reference Framework reached version 2.9.0 on May 21, 2026, and continues to be accompanied by implementing acts, specifications, and reference implementations. Principal references: Regulation (EU) 2016/679 (GDPR) [SRC-S34-005]; Regulation (EU) 2024/1183 on European digital identity [SRC-S34-008]; European Commission, EUDI Wallet implementation and ARF v2.9.0 [SRC-S34-009]; NIST SP 800-63-4 (2025) [SRC-NIST-ID4].

Chapter 4

• 7. Automated decision-making and the transformation of work. Article 22 of the GDPR concerns decisions based solely on automated processing that produce legal or similarly significant effects. As of July 17, 2026, the prohibitions, AI literacy obligations, and rules for general-purpose AI models are applicable. For GPAI models, the obligations under Article 53 have applied since August 2, 2025; the final code of July 10, 2025, is the current version and, under the timetable indicated by the Commission, enforcement by the AI Office is scheduled to begin on August 2, 2026, for new models and August 2, 2027, for models already on the market. The Platform Work Directive must be transposed by December 2, 2026, a deadline that had not yet passed at the time of verification. The 2025 ILO index measures occupational exposure and identifies role transformation as the most likely outcome; it is not a forecast of jobs eliminated. The threshold adopted in this book is deliberately broader. Principal references: Regulation (EU) 2024/1689 (AI Act) [SRC-S34-001]; European Commission—AI Act Service Desk, Article 53 [SRC-S34-004]; European Commission, General-Purpose AI Code of Practice (2025) [SRC-S34-021]; Regulation (EU) 2016/679 (GDPR) [SRC-S34-005]; EDPB, guidelines on automated decision-making and profiling [SRC-S34-006]; CJEU, SCHUFA, C-634/21 (2023) [SRC-S34-007]; Directive (EU) 2024/2831 on platform work [SRC-S34-012]; ILO, Working Paper 140 (2025) [SRC-ILO-GENAI-JOBS25].

Chapters 5–6

• 8. Urban biometrics and education. EU law prohibits or restricts certain biometric uses, especially in law enforcement, but does not establish a uniform ban on every form of urban facial recognition. The formulation in Chapter 5 is a policy threshold proposed by the author. In education, UNESCO recommends human-centered governance, data protection, pedagogical validation, and age-appropriate approaches; the future institutions in Chapter 6 are scenarios, not predictions. Principal references: Regulation (EU) 2024/1689 (AI Act) [SRC-S34-001]; AI Act, Article 5 [SRC-S34-011]; EDPB, guidelines on facial recognition in law enforcement [SRC-S34-017]; UNESCO, guidance on generative AI in education (2023) [SRC-UNESCO-GENAI-ED23].

Chapter 8

• 9. AI in health care and mental health. The AI Act and the regulations on medical devices and in vitro diagnostic medical devices interact according to function, intended purpose, and classification. Under MDCG 2019-11 rev.1, software qualification depends on its stated intended purpose, not on whether it operates on a device, in the cloud, or on another platform. The joint AIB 2025-1/MDCG 2025-6 guidance further clarifies that classification as high-risk under the AI Act does not, by itself, change the software’s MDR or IVDR class. Clinical responsibility also depends on national professional law; the nondelegability proposed in this book is not presented as a uniform rule currently in force. WHO calls for the assessment of risks and benefits, data quality, transparency, monitoring, and accountability throughout the life cycle; institutional documentation does not establish universal therapeutic efficacy for conversational tools. In 2026, WHO added a report on the adoption of AI in health systems across the 27 EU Member States and a discussion paper on AI in evidence-informed health policy: both reinforce the need for governance, oversight, and human judgment without replacing the earlier clinical and regulatory guidance. Principal references: Regulation (EU) 2024/1689 (AI Act) [SRC-S34-001]; Regulation (EU) 2017/745 on medical devices [SRC-S34-015]; Regulation (EU) 2017/746 on in vitro diagnostic medical devices [SRC-S34-016]; MDCG 2019-11 rev.1 (2025) [SRC-MDCG-MDSW25]; AIB 2025-1/MDCG 2025-6 [SRC-MDCG-AIA25]; WHO, ethics and governance of AI for health (2021) [SRC-WHO-AI-HEALTH21]; WHO, regulatory considerations on AI for health (2023) [SRC-WHO-AI-REG23]; WHO, guidance on large multimodal models for health (2024 edition; WHO page dated 2025) [SRC-WHO-LMM24-25]; WHO/Europe, health-system preparedness for AI in the EU (2026) [SRC-WHO-AI-EU26]; WHO, AI and evidence-informed health policy (2026) [SRC-WHO-AI-POLICY26].

Chapter 9

• 10. Synthetic content. Beginning August 2, 2026, Article 50 of the AI Act provides for transparency obligations covering certain interactions and categories of synthetic content or deepfakes. The final EU code of June 10, 2026, is voluntary and provides guidance on marking and labeling. The Commission’s opinion of July 8, 2026, and the AI Board’s assessment of July 9 consider it an adequate tool for facilitating compliance, but adherence alone does not demonstrate compliance with Article 50. Neither marking nor provenance is equivalent to truth or creates a universal procedure for contesting content. The attribution and definition of the “liar’s dividend” follow Robert Chesney and Danielle Keats Citron (2019). Principal references: Regulation (EU) 2024/1689 (AI Act) [SRC-S34-001]; AI Act, Article 50 [SRC-S34-003]; European Commission, Code of Practice on Transparency of AI-Generated Content and adequacy assessments (2026) [SRC-S34-020]; Chesney and Citron (2019) [SRC-LIARS-DIVIDEND].

Chapter 10

• 11. Copyright and training. The DSM Directive distinguishes between text and data mining exceptions and the reservation of rights; since August 2, 2025, the AI Act has imposed obligations on providers of general-purpose AI models that include a policy for complying with Union copyright law and a sufficiently detailed summary of the content used to train the model, subject to the transitions provided for models already placed on the market. Licenses, consent, exceptions, and litigation remain variable, and style is not treated as a uniform property right. Principal references: Regulation (EU) 2024/1689 (AI Act) [SRC-S34-001]; AI Act, Article 53 [SRC-S34-004]; Directive (EU) 2019/790 on copyright [SRC-S34-013].

Chapter 11

• 12. Minors, loneliness, and artificial companions. The AI Act and the Digital Services Act protect minors in defined circumstances, but there is no single rule governing every form of simulated intimacy. WHO documents the vulnerability of adolescent mental health and treats loneliness and isolation as public health concerns; UNICEF sets out safeguards grounded in children’s rights. UNICEF’s June 2026 policy brief on AI companion chatbots identifies specific risks and regulatory gaps, but also calls for further research on benefits and harms: it is policy guidance, not clinical evidence. The FTC’s 2025 inquiry demonstrates regulatory attention; by itself, it does not establish clinical effects. The risks of attachment, dependency, and displacement of human relationships are plausible and are discussed in available guidance and reviews, but they are neither inevitable outcomes nor already established causal effects. The limit on love, exclusivity, jealousy, and abandonment is the author’s proposal. Principal references: Regulation (EU) 2024/1689 (AI Act) [SRC-S34-001]; Regulation (EU) 2022/2065 (Digital Services Act) [SRC-S34-010]; AI Act, Article 5 [SRC-S34-011]; WHO, adolescent mental health (2025) [SRC-WHO-ADOLESCENT25]; WHO, Commission on Social Connection (2025) [SRC-WHO-CONNECTION25]; UNICEF, Guidance on AI and Children 3.0 (2025) [SRC-UNICEF-AI-CHILD25]; UNICEF, When AI becomes a friend (2026) [SRC-UNICEF-COMPANIONS26]; FTC, inquiry into AI companion chatbots (2025) [SRC-FTC-COMPANIONS25].

Chapter 12

• 13. Profiling, contestation, human intervention, and fairness audits. Information, rectification, objection, and human intervention have different legal bases and scopes under the GDPR; the chapter explicitly brings them together in a proposed civil threshold. Article 10(5) of the AI Act allows providers of high-risk systems, under strict conditions, to process special categories of personal data exceptionally when this is strictly necessary to detect and correct bias: it is not a general authorization for fairness audits. Principal references: Regulation (EU) 2016/679 (GDPR) [SRC-S34-005]; EDPB, guidelines on automated decision-making and profiling [SRC-S34-006]; CJEU, SCHUFA, C-634/21 (2023) [SRC-S34-007]; Regulation (EU) 2024/1689 (AI Act), Article 10(5) [SRC-S34-001].

Chapter 13

• 14. Energy, water, materials, and electronic waste. Statements about data centers, electricity demand, water use, critical materials, and end of life are supported by reports from the IEA, Berkeley Lab, USGS, and the United Nations; the body of the book avoids presenting fixed quantities as real-world data and attributes impacts to supply chains, location, and the life cycle. In USGS terminology, withdrawal is water removed from a source, whereas consumption is the portion not immediately returned to the local environment. ISO 14040 and ISO 14044 require a life cycle assessment to include goal and scope definition, inventory analysis, impact assessment, and interpretation. These methodological boundaries do not substitute for site-specific data: any future quantification will need to identify the source, site, season, functional unit, and system boundaries. Principal references: IEA, Energy and AI (2025) [SRC-IEA-EAI25]; Lawrence Berkeley National Laboratory, Data Center Energy Usage Report (2024) [SRC-LBNL-DC24]; USGS, Water-Use Terminology [SRC-USGS-WATER]; ISO 14040:2006 and ISO 14044:2006 [SRC-ISO-LCA14040-44]; USGS, Mineral Commodity Summaries (2026) [SRC-USGS-MCS26]; IEA, Global Critical Minerals Outlook (2025) [SRC-IEA-CM25]; ITU and UNITAR, Global E-waste Monitor (2024) [SRC-UNITAR-EW24]. The values in the city project table—capacity, hours, energy, water, materials, and costs—are parameters constructed for the scenario: they do not come from a real facility, are not market benchmarks, and are not offered as a forecast. They illustrate a method of comparison when units, season, and boundaries are stated.
• 15. Orbital computing: demonstrations, announcements, and scenario. Three documentary categories were distinguished in developing the passage on orbital computing: demonstrations already launched, future announcements, and the 2050 scenario. The Starcloud and Axiom pages document only their own activities; the Google Research page documents an announced project, not a completed launch. The constraints posed by exposure and eclipse, storage, radiative heat rejection, mass, reliability, maintenance, communications, service life, and the life cycle are verified at two levels: ASCEND/CORDIS as an institutional feasibility study, not an independent source, and independent technical literature. Principal references: Starcloud, Starcloud-1 [SRC-STARCLOUD1-25]; Axiom Space, Orbital Data Centers [SRC-AXIOM-ODC26]; Google Research, Project Suncatcher (2025) [SRC-GOOGLE-SUNCATCHER25]; European Commission, CORDIS ASCEND [SRC-CORDIS-ASCEND]; Turyshev (2026) [SRC-TURYSHEV-ODC26]; Al Ahmad, Memon, and Pecht (2026) [SRC-ALAHMAD-SBDC26]; Wilson et al. (2023) [SRC-WILSON-LCSA23]; Murphy et al. (2023) [SRC-MURPHY-REENTRY23].

Chapter 14

• 16. Supply chains and semiconductors. Supply-chain specialization is documented by NIST and GAO; UNCTAD provides the broader framework for AI capacity, distribution, and inclusive development. With respect to U.S. controls, the BIS final rule of January 15, 2026 (91 FR 1684; 15 CFR parts 742, 744, and 748), changes the license review policy from a presumption of denial to case-by-case review for certain advanced computing commodities exported from the United States to end users in China or Macao, when specific conditions are met; it does not regulate all semiconductors in the same way. Rule 91 FR 17851, published April 9, 2026, and effective April 7, extends the transitional status of certain Authorized IC Designers through December 31, 2026, without replacing the January policy; the BIS FAQs of May 2026 provide operational guidance. Scope and requirements must be checked against the EAR then in force and the specific transaction. The European Chips Act 2.0 remains a June 2026 Commission proposal, not law already enacted. Principal references: NIST, semiconductor supply chain (2023) [SRC-NIST-CHIPS-SUPPLY]; GAO, semiconductor supply chain (2022) [SRC-GAO-SEMI22]; European Commission, European Chips Act [SRC-EU-CHIPS]; European Commission, Chips Act 2.0 proposal (2026) [SRC-EU-CHIPS2-26]; Bureau of Industry and Security, semiconductor rules and guidance (2026) [SRC-BIS-AC26]; UNCTAD, Technology and Innovation Report (2025) [SRC-UNCTAD-TIR25].
• 17. Cables, cloud, and cybersecurity. The CMA’s final decision of July 31, 2025, on the United Kingdom’s public cloud infrastructure market finds adverse effects on competition and recommends considering SMS investigations into Microsoft and AWS. In March 2026, however, the CMA Board chose a package of actions that includes monitoring commitments on egress fees and interoperability and an SMS investigation into Microsoft’s software ecosystem, launched on May 14, 2026. These developments do not amount to an SMS designation or demonstrate abuse in every market. The final report of the International Advisory Body on Submarine Cable Resilience, established jointly by ITU and ICPC, was approved on July 10, 2026, and addresses cable resilience, repair, risk mitigation, and geographic diversity. ENISA Threat Landscape 2025 remains the most recent annual edition; ENISA’s July 7, 2026, report on cybersecurity in the frontier AI era adds initial recommendations for countering machine-speed threats and is not a new edition of the Threat Landscape. The chapter’s geopolitical configurations remain conditional scenarios, not predictions. Principal references: ITU and ICPC, report on submarine cable resilience (2026) [SRC-ITU-CABLES26]; CMA, final decision on the cloud market (2025) and 2026 actions [SRC-CMA-CLOUD25]; ENISA, Threat Landscape (2025) [SRC-ENISA-TL25] and Cybersecurity in the Frontier AI Era (2026) [SRC-ENISA-FRONTIER26]; United Nations, OEWG report on ICT security (2021) [SRC-UN-CYBER].
• 18. Mombasa and the Northern Corridor. For the geographic context of Mombasa, the Kenya Ports Authority documents the port’s multimodal connection to its hinterland; the Northern Corridor Transit and Transport Coordination Authority documents the multimodal route between the Port of Mombasa and the landlocked countries of the Great Lakes region. No quantitative data on capacity, distances, or transit times are carried into the text. Principal references: Kenya Ports Authority, Port of Mombasa [SRC-KPA-MOMBASA]; Northern Corridor Transit and Transport Coordination Authority, Who We Are [SRC-NCTTCA-NORTHERN-CORRIDOR].

Chapter 15

• 19. Data relating to deceased persons, grief, and the digital afterlife. The GDPR does not apply to the personal data of deceased persons and permits national rules. In Italy, Article 2-terdecies of the Privacy Code allows specified persons to exercise certain rights, subject to limitations. The cited ethical literature frames responsibilities and the posthumous handling of data, but does not provide a clinical basis for asserting universal benefits or harms from griefbots. The book’s post-mortem digital directive remains a proposal. Principal references: Regulation (EU) 2016/679 (GDPR) [SRC-S34-005]; Italian Data Protection Authority, digital inheritance and Article 2-terdecies [SRC-S34-018]; Öhman and Floridi (2018) [SRC-OHMAN-AFTERLIFE18].

Essential Timeline

• 20. ARPANET, TCP/IP, and the World Wide Web. The milestones in the Essential Timeline are documented by DARPA, the Internet Society, and CERN. Principal references: DARPA, ARPANET [SRC-DARPA-ARPANET-1969]; Internet Society, TCP/IP [SRC-ISOC-TCPIP-1983]; CERN, history of the Web [SRC-CERN-WEB-HISTORY]; CERN, opening of the Web [SRC-CERN-WEB-1991].

Documentary Bibliography

This section gathers the sources actually used to verify the book’s factual and legal claims. Each entry identifies, as applicable, a primary source, standard, legislative act, proposal, interpretive guidance, institutional report, scientific article, preprint, or corporate source; version, date, and status are provided when relevant. The [SRC-…] code is a stable technical identifier and also appears in the Endnotes alongside the readable reference.

History of Artificial Intelligence and Networks

• [SRC-TURING-1950] A. M. Turing, Computing Machinery and Intelligence. Oxford University Press / Mind. 1950. Primary source. https://doi.org/10.1093/mind/LIX.236.433.
• [SRC-DARTMOUTH-1955] John McCarthy, Marvin L. Minsky, Nathaniel Rochester, and Claude E. Shannon, A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence. Dartmouth College archive. 1955. Primary source / official archive. https://jmc.stanford.edu/articles/dartmouth/dartmouth.pdf.
• [SRC-ROSENBLATT-1958] Frank Rosenblatt, The Perceptron: A Probabilistic Model for Information Storage and Organization in the Brain. Psychological Review. 1958. Primary source. https://doi.org/10.1037/h0042519.
• [SRC-MINSKY-PAPERT-1969] Marvin Minsky and Seymour Papert, Perceptrons: An Introduction to Computational Geometry. MIT Press, 1969; expanded edition, 1987. Primary source; the publisher’s page refers to the expanded edition. https://mitpress.mit.edu/9780262631112/perceptrons/.
• [SRC-RHW-1986] David E. Rumelhart, Geoffrey E. Hinton, and Ronald J. Williams, Learning representations by back-propagating errors. Nature. 1986. Primary source. https://doi.org/10.1038/323533a0.
• [SRC-IBM-DEEPBLUE] IBM, Deep Blue. 1997. Official corporate archive. Accessed: 2026. https://www.ibm.com/history/deep-blue.
• [SRC-ALEXNET-2012] Alex Krizhevsky, Ilya Sutskever, and Geoffrey E. Hinton, ImageNet Classification with Deep Convolutional Neural Networks. NeurIPS. 2012. Primary source. https://papers.nips.cc/paper/4824-imagenet-classification-with-deep-convolutional-neural-networks.
• [SRC-TRANSFORMER-2017] Ashish Vaswani et al., Attention Is All You Need. NeurIPS / arXiv. 2017. Primary source. https://arxiv.org/abs/1706.03762.
• [SRC-ELIZA-1966] Joseph Weizenbaum, ELIZA—A Computer Program for the Study of Natural Language Communication Between Man and Machine. Communications of the ACM. 1966. Primary source. https://doi.org/10.1145/365153.365168.
• [SRC-OPENAI-CHATGPT] OpenAI, Introducing ChatGPT. 2022-11-30. Official corporate announcement. https://openai.com/index/chatgpt/.
• [SRC-DARPA-ARPANET-1969] DARPA, ARPANET. 1969. Official institutional archive. Accessed: 2026. https://www.darpa.mil/news/features/arpanet.
• [SRC-ISOC-TCPIP-1983] Internet Society, New Year’s Day Marks 30th Anniversary of Major Milestone for Global Internet. 2013-01-01. Institutional documentation. https://www.internetsociety.org/news/press-releases/2013/new-years-day-marks-30th-anniversary-of-major-milestone-for-global-internet/.
• [SRC-CERN-WEB-HISTORY] CERN, A short history of the Web. Official institutional archive. Accessed: 2026. https://home.cern/science/computing/the-birth-of-the-web/short-history-web/.
• [SRC-CERN-WEB-1991] CERN, The open internet and the web. 2013-04-30. Official documentation. https://home.cern/open-internet-and-web/.

Digital Identity, Credentials, and Standards

• [SRC-W3C-VC20] W3C Verifiable Credentials Working Group, Verifiable Credentials Data Model v2.0. 2025-05-15. W3C Recommendation. https://www.w3.org/TR/vc-data-model-2.0/.
• [SRC-W3C-DI] W3C Verifiable Credentials Working Group, Verifiable Credential Data Integrity 1.0. 2025-05-15. W3C Recommendation. https://www.w3.org/TR/vc-data-integrity/.
• [SRC-W3C-JOSE] W3C Verifiable Credentials Working Group, Securing Verifiable Credentials using JOSE and COSE. 2025-05-15. W3C Recommendation. https://www.w3.org/TR/vc-jose-cose/.
• [SRC-OID4VCI] OpenID Foundation, OpenID for Verifiable Credential Issuance 1.0. 2025-09-16. Final Specification. https://openid.net/specs/openid-4-verifiable-credential-issuance-1_0.html.
• [SRC-OID4VP] OpenID Foundation, OpenID for Verifiable Presentations 1.0. 2025-07-09. Final Specification. https://openid.net/specs/openid-4-verifiable-presentations-1_0.html.
• [SRC-NIST-ID4] National Institute of Standards and Technology, Digital Identity Guidelines, NIST SP 800-63-4. 2025-07-31. Final technical standard. https://doi.org/10.6028/NIST.SP.800-63-4.
• [SRC-NIST-8202] Dylan Yaga et al., Blockchain Technology Overview, NISTIR 8202. 2018-10-03. Official technical report. https://doi.org/10.6028/NIST.IR.8202.

Health, Society, Education, Work, and Development

• [SRC-WHO-AI-HEALTH21] World Health Organization, Ethics and governance of artificial intelligence for health. 2021-06-28. Institutional guidance. https://www.who.int/publications/i/item/9789240029200.
• [SRC-WHO-AI-REG23] World Health Organization, Regulatory considerations on artificial intelligence for health. 2023-10-19. Institutional framework of regulatory considerations, not a regulation. https://www.who.int/publications/i/item/9789240078871.
• [SRC-WHO-LMM24-25] World Health Organization, Ethics and governance of artificial intelligence for health: Guidance on large multi-modal models. WHO edition and bibliographic citation: 2024; current institutional page dated 2025-03-25. https://www.who.int/publications/i/item/9789240084759; https://www.who.int/publications/b/70584.
• [SRC-WHO-AI-EU26] WHO Regional Office for Europe, Artificial intelligence is reshaping health systems: state of readiness across the European Union. 2026-04-20. Institutional report on the preparedness of health systems in the 27 EU Member States; not clinical evidence of efficacy. WHO/EURO:2026-12707-52481-81471. https://www.who.int/europe/publications/i/item/WHO-EURO-2026-12707-52481-81471.
• [SRC-WHO-AI-POLICY26] World Health Organization, Artificial intelligence and evidence-informed policy: emerging challenges and opportunities: discussion paper. 2026-04-25. Institutional technical paper; calls for oversight, impact assessments, and human judgment. https://www.who.int/publications/i/item/B09667.
• [SRC-MDCG-MDSW25] Medical Device Coordination Group, MDCG 2019-11 rev.1, Guidance on Qualification and Classification of Software in Regulation (EU) 2017/745 — MDR and Regulation (EU) 2017/746 — IVDR. 2025-06-17. Official European Commission guidance, not a legislative act. https://health.ec.europa.eu/document/download/b45335c5-1679-4c71-a91c-fc7a4d37f12b_en.
• [SRC-MDCG-AIA25] Artificial Intelligence Board / Medical Device Coordination Group, AIB 2025-1 — MDCG 2025-6, Interplay between the Medical Devices Regulation, the In Vitro Diagnostic Medical Devices Regulation and the Artificial Intelligence Act. 2025. Official joint guidance, not a legislative act. https://health.ec.europa.eu/document/download/b78a17d7-e3cd-4943-851d-e02a2f22bbb4_en.
• [SRC-WHO-ADOLESCENT25] World Health Organization, Mental health of adolescents. Updated 2025-09-01. Institutional fact sheet. https://www.who.int/news-room/fact-sheets/detail/adolescent-mental-health.
• [SRC-WHO-CONNECTION25] World Health Organization, From loneliness to social connection: charting a path to healthier societies — Report of the WHO Commission on Social Connection. 2025-06-30. Institutional report. https://www.who.int/groups/commission-on-social-connection/report.
• [SRC-UNICEF-AI-CHILD25] UNICEF Innocenti, Guidance on AI and Children, Version 3.0. 2025-12. Institutional guidance grounded in children’s rights. https://www.unicef.org/innocenti/reports/policy-guidance-ai-children.
• [SRC-UNICEF-COMPANIONS26] UNICEF, When AI becomes a friend: Child rights risks, harms, and regulatory responses to AI chatbots and companions. 2026-06. Policy brief; regulatory review updated through 2026-05-15, not clinical evidence. https://www.unicef.org/documents/when-ai-becomes-friend-child-rights-risks.
• [SRC-FTC-COMPANIONS25] U.S. Federal Trade Commission, FTC Launches Inquiry into AI Chatbots Acting as Companions. 2025-09-11. Regulatory inquiry, not a clinical finding. https://www.ftc.gov/news-events/news/press-releases/2025/09/ftc-launches-inquiry-ai-chatbots-acting-companions.
• [SRC-UNESCO-GENAI-ED23] UNESCO, Guidance for generative AI in education and research. 2023-09-07, page updated 2026-01-16. Institutional guidance. https://www.unesco.org/en/articles/guidance-generative-ai-education-and-research.
• [SRC-ILO-GENAI-JOBS25] Paweł Gmyrek et al., Generative AI and Jobs: A Refined Global Index of Occupational Exposure. ILO Working Paper 140. 2025-05-20. Institutional research report. https://doi.org/10.54394/HETP0387.
• [SRC-OHMAN-AFTERLIFE18] Carl Öhman and Luciano Floridi, An ethical framework for the digital afterlife industry. Nature Human Behaviour, 2, 318–320. 2018-04-09. Scientific article. https://doi.org/10.1038/s41562-018-0335-2.
• [SRC-UNCTAD-TIR25] UN Trade and Development, Technology and Innovation Report 2025: Inclusive artificial intelligence for development. 2025-04-07. UN report. https://unctad.org/publication/technology-and-innovation-report-2025.

Energy, Materials, and the Geopolitics of Infrastructure

• [SRC-IEA-EAI25] International Energy Agency (IEA), Energy and AI. 2025-04-10. Institutional report. https://www.iea.org/reports/energy-and-ai.
• [SRC-LBNL-DC24] Arman Shehabi et al., 2024 United States Data Center Energy Usage Report. Lawrence Berkeley National Laboratory / U.S. Department of Energy, 2024-12-19. Official technical report. DOI 10.71468/P1WC7Q. https://eta.lbl.gov/publications/2024-lbnl-data-center-energy-usage-report.
• [SRC-USGS-WATER] U.S. Geological Survey, Water-Use Terminology. Official page. Accessed: 2026-07-15. Operational definitions of withdrawal and consumptive use. https://www.usgs.gov/mission-areas/water-resources/science/water-use-terminology.
• [SRC-ISO-LCA14040-44] International Organization for Standardization, ISO 14040:2006, Environmental management — Life cycle assessment — Principles and framework; ISO 14044:2006, Environmental management — Life cycle assessment — Requirements and guidelines. Current editions confirmed in 2022. Methodological standards. https://www.iso.org/standard/37456.html; https://www.iso.org/standard/38498.html.
• [SRC-USGS-MCS26] U.S. Geological Survey, Mineral Commodity Summaries 2026. Version 1.3, May 2026. Official report. https://doi.org/10.3133/mcs2026.
• [SRC-IEA-CM25] International Energy Agency (IEA), Global Critical Minerals Outlook 2025. 2025. Institutional report. https://www.iea.org/reports/global-critical-minerals-outlook-2025.
• [SRC-UNITAR-EW24] UNITAR SCYCLE and International Telecommunication Union, Global E-waste Monitor 2024. 2024-03-20. UN report. https://www.itu.int/hub/publication/d-gen-e_waste-01-2024/.
• [SRC-STARCLOUD1-25] Starcloud, Starcloud-1. Corporate page about its own activities; states that the satellite was launched in November 2025, without demonstrating scale or commercial viability. Accessed: 2026-07-17. https://www.starcloud.com/starcloud-1.
• [SRC-AXIOM-ODC26] Axiom Space, Orbital Data Centers. Corporate page about its own activities; states that the first two dedicated nodes were launched into low Earth orbit on January 11, 2026. Accessed: 2026-07-17. https://www.axiomspace.com/orbital-data-center.
• [SRC-GOOGLE-SUNCATCHER25] Google Research, Exploring a space-based, scalable AI infrastructure system design. 2025-11-04. Corporate announcement of Project Suncatcher; the two prototype satellites planned for early 2027 remain an announced project, not a completed launch. Accessed: 2026-07-17. https://research.google/blog/exploring-a-space-based-scalable-ai-infrastructure-system-design/.
• [SRC-CORDIS-ASCEND] European Commission, CORDIS, Advanced Space Cloud for European Net zero emissions and Data sovereignty (ASCEND), grant agreement 101082517: Fact Sheet and Periodic Reporting. Project: 2023-01-01–2024-04-30; report updated: 2026-01-14. Institutional feasibility fact sheet and report; not an independent technical source. Project DOI 10.3030/101082517. Accessed: 2026-07-17. https://cordis.europa.eu/project/id/101082517; https://cordis.europa.eu/project/id/101082517/reporting.
• [SRC-TURYSHEV-ODC26] Slava G. Turyshev, Orbital Data Centers: Spacecraft Constraints and Economic Viability. 2026-04-29. Independent technical preprint, arXiv:2604.27197. DOI 10.48550/arXiv.2604.27197. Accessed: 2026-07-17. https://arxiv.org/abs/2604.27197; https://doi.org/10.48550/arXiv.2604.27197.
• [SRC-ALAHMAD-SBDC26] Mahmoud Al Ahmad, Qurban Memon, and Michael Pecht, Reliability and Risk in Space-Based Data Centers: A Lifecycle Assessment of Orbital Cloud Infrastructure. Applied Sciences, 16(11), 5247. 2026-05-23. Independent peer-reviewed technical article. DOI 10.3390/app16115247. Accessed: 2026-07-17. https://www.mdpi.com/2076-3417/16/11/5247; https://doi.org/10.3390/app16115247.
• [SRC-WILSON-LCSA23] Andrew R. Wilson, Massimiliano Vasile, Christie Maddock, and Keith Baker, Implementing life cycle sustainability assessment for improved space mission design. Integrated Environmental Assessment and Management, 19, 1002–1022, 2023. Independent peer-reviewed technical article. DOI 10.1002/ieam.4722. Accessed: 2026-07-17. https://onlinelibrary.wiley.com/doi/10.1002/ieam.4722; https://doi.org/10.1002/ieam.4722.
• [SRC-MURPHY-REENTRY23] Daniel M. Murphy et al., Metals from spacecraft reentry in stratospheric aerosol particles. Proceedings of the National Academy of Sciences, 120(43), 2023. Independent peer-reviewed scientific article; copy in NOAA’s institutional repository. DOI 10.1073/pnas.2313374120. Accessed: 2026-07-17. https://repository.library.noaa.gov/view/noaa/58476; https://doi.org/10.1073/pnas.2313374120.
• [SRC-KPA-MOMBASA] Kenya Ports Authority, Port of Mombasa. Institutional page on the port’s multimodal connections with its hinterland. Accessed: 2026-07-17. https://kpa.co.ke/Ports/PortOfMombasa.
• [SRC-NCTTCA-NORTHERN-CORRIDOR] Northern Corridor Transit and Transport Coordination Authority, Who We Are. Institutional page on the Northern Corridor between the Port of Mombasa and the landlocked countries of the Great Lakes region. Accessed: 2026-07-17. https://www.ttcanc.org/node/70.
• [SRC-NIST-CHIPS-SUPPLY] NIST, Vision for Success: Facilities for Semiconductor Materials and Manufacturing Equipment. NIST / CHIPS for America. 2023-06-22. Official documentation. https://www.nist.gov/chips/vision-success-facilities-semiconductor-materials-and-manufacturing-equipment.
• [SRC-GAO-SEMI22] U.S. Government Accountability Office, Semiconductor Supply Chain: Policy Considerations from Selected Experts for Reducing Risks and Mitigating Shortages. 2022-07-26. Official report. https://www.gao.gov/products/gao-22-105923.
• [SRC-EU-CHIPS] European Commission, European Chips Act. Institutional page on the current framework; not a legislative act. Accessed: 2026. https://digital-strategy.ec.europa.eu/en/policies/european-chips-act.
• [SRC-EU-CHIPS2-26] European Commission, Proposal for the Chips Act 2.0. 2026-06-03. Legislative proposal, not law already enacted. https://digital-strategy.ec.europa.eu/en/library/proposal-chips-act-20.
• [SRC-BIS-AC26] U.S. Bureau of Industry and Security, Revision to License Review Policy for Advanced Computing Commodities. Final rule, 91 FR 1684, 15 CFR parts 742, 744, and 748, published and effective 2026-01-15; Extension of Authorized Integrated Circuit Designer Status, 91 FR 17851, published 2026-04-09 and effective 2026-04-07; Guidance on Advanced Computing Items, FAQ dated 2026-05. Official acts and guidance; applicability must be verified against the EAR in force and the specific transaction. Accessed: 2026-07-17. https://www.federalregister.gov/documents/2026/01/15/2026-00789/revision-to-license-review-policy-for-advanced-computing-commodities; https://www.bis.gov/regulations/federal-register-notices; https://www.bis.gov/regulations/ear/744; https://www.bis.gov/regulations/ear/748.
• [SRC-ITU-CABLES26] International Telecommunication Union and International Cable Protection Committee, International Advisory Body on Submarine Cable Resilience — Report of the Working Groups 2026. Approved 2026-07-10. Final institutional report. https://www.itu.int/hub/publication/s-iab-wg-2026/.
• [SRC-CMA-CLOUD25] UK Competition and Markets Authority, Cloud services market investigation — final decision. 2025-07-31. Final regulatory decision on the UK public cloud infrastructure market; update: Actions on cloud and business software through the UK digital markets competition regime, 2026-03-31, and launch of the SMS investigation into Microsoft’s software ecosystem, 2026-05-14. https://www.gov.uk/cma-cases/cloud-services-market-investigation; https://www.gov.uk/government/publications/business-software-and-cloud-services.
• [SRC-ENISA-TL25] ENISA, ENISA Threat Landscape 2025. 2025-10-01. Official report. https://www.enisa.europa.eu/publications/enisa-threat-landscape-2025.
• [SRC-ENISA-FRONTIER26] ENISA, ENISA’s view on Cybersecurity in the Frontier AI Era. 2026-07-07. Official report with initial recommendations for operational capabilities against machine-speed threats; does not replace the annual Threat Landscape. https://www.enisa.europa.eu/publications/enisas-view-on-cybersecurity-in-the-frontier-ai-era.
• [SRC-UN-CYBER] United Nations, Final substantive report of the Open-ended Working Group on developments in the field of information and telecommunications in the context of international security, A/75/816. 2021-03-18. Official UN report. https://digitallibrary.un.org/record/3908015.

Law, Regulation, and Institutional Guidance

• [SRC-LIARS-DIVIDEND] Robert Chesney and Danielle Keats Citron, Deep Fakes: A Looming Challenge for Privacy, Democracy, and National Security. California Law Review, vol. 107, 2019, pp. 1753–1819. Primary source for the concept of the “liar’s dividend”; DOI 10.15779/Z38RV0D15J. https://www.californialawreview.org/print/deep-fakes-a-looming-challenge-for-privacy-democracy-and-national-security.
• [SRC-S34-001] European Parliament and Council, Regulation (EU) 2024/1689 (AI Act). 2024-06-13. Law in force; phased application. CELEX 32024R1689; ELI reg/2024/1689/oj. https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng.
• [SRC-S34-003] European Commission—AI Act Service Desk, AI Act—Article 50: Transparency obligations for providers and deployers of certain AI systems. Official service concerning Article 50 of Regulation (EU) 2024/1689. Accessed: 2026-07-17. https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-50.
• [SRC-S34-004] European Commission—AI Act Service Desk, AI Act—Article 53: Obligations for providers of general-purpose AI models. Official service concerning Article 53 of Regulation (EU) 2024/1689. Accessed: 2026-07-17. https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-53.
• [SRC-S34-005] European Parliament and Council, Regulation (EU) 2016/679 (GDPR). 2016-04-27. Law in force. CELEX 32016R0679; ELI reg/2016/679/oj. https://eur-lex.europa.eu/eli/reg/2016/679/oj/eng.
• [SRC-S34-006] Article 29 Data Protection Working Party, Guidelines on Automated individual decision-making and Profiling for the purposes of Regulation 2016/679, WP251 rev.01. Endorsed by the European Data Protection Board on 2018-05-25. Official interpretive guidance; not legislation. https://www.edpb.europa.eu/documents/guideline/automated-decision-making-and-profiling_en.
• [SRC-S34-007] Court of Justice of the European Union, SCHUFA Holding (Scoring), C-634/21. 2023-12-07. Binding case law in the interpretation of EU law. ECLI:EU:C:2023:957. https://infocuria.curia.europa.eu/tabs/redirect/juris/document/document.jsf?docid=280426&doclang=en.
• [SRC-S34-008] European Parliament and Council, Regulation (EU) 2024/1183 establishing the European Digital Identity Framework. 2024-04-11. Law in force. CELEX 32024R1183; ELI reg/2024/1183/oj. https://eur-lex.europa.eu/eli/reg/2024/1183/oj/eng.
• [SRC-S34-009] European Commission, European Digital Identity Wallet implementation. Page updated: 2026-06-22. The framework provides for Member State wallets by the end of 2026; the official Architecture and Reference Framework repository identifies v2.9.0, dated 2026-05-21, as the current release. Institutional page and technical specification. Accessed: 2026-07-17. https://digital-strategy.ec.europa.eu/en/policies/eudi-wallet-implementation; https://github.com/eu-digital-identity-wallet/eudi-doc-architecture-and-reference-framework/releases/tag/v2.9.0.
• [SRC-S34-010] European Parliament and Council, Regulation (EU) 2022/2065 on digital services (DSA). 2022-10-19. Law in force. CELEX 32022R2065; ELI reg/2022/2065/oj. https://eur-lex.europa.eu/eli/reg/2022/2065/oj/eng.
• [SRC-S34-011] European Commission—AI Act Service Desk, AI Act—Article 5: Prohibited AI practices. Official service concerning Article 5 of Regulation (EU) 2024/1689. Accessed: 2026-07-14. https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-5.
• [SRC-S34-012] European Parliament and Council, Directive (EU) 2024/2831 on improving working conditions in platform work. 2024-10-23. EU law requiring transposition; sector-specific scope. CELEX 32024L2831; ELI dir/2024/2831/oj. https://eur-lex.europa.eu/eli/dir/2024/2831/oj/eng.
• [SRC-S34-013] European Parliament and Council, Directive (EU) 2019/790 on copyright in the Digital Single Market. 2019-04-17. EU law transposed with national differences. CELEX 32019L0790; ELI dir/2019/790/oj. https://eur-lex.europa.eu/eli/dir/2019/790/oj/eng.
• [SRC-S34-015] European Parliament and Council, Regulation (EU) 2017/745 on medical devices. 2017-04-05. Law in force. CELEX 32017R0745; ELI reg/2017/745/oj. https://eur-lex.europa.eu/eli/reg/2017/745/oj/eng.
• [SRC-S34-016] European Parliament and Council, Regulation (EU) 2017/746 on in vitro diagnostic medical devices. 2017-04-05. Law in force. CELEX 32017R0746; ELI reg/2017/746/oj. https://eur-lex.europa.eu/eli/reg/2017/746/oj/eng.
• [SRC-S34-017] European Data Protection Board, Guidelines 05/2022 on the use of facial recognition technology in the area of law enforcement. Final version 2.0, 2023-05-17. Official interpretive guidance; not law. https://www.edpb.europa.eu/system/files/2023-05/edpb_guidelines_202304_frtlawenforcement_v2_en.pdf.
• [SRC-S34-018] Italian Data Protection Authority, Eredità digitale e art. 2-terdecies del Codice privacy. Institutional guidance; legal basis: Legislative Decree 196/2003, Article 2-terdecies. Accessed: 2026-07-14. https://www.garanteprivacy.it/temi/internet-e-nuove-tecnologie/eredita-digitale.
• [SRC-S34-020] European Commission, Code of Practice on Transparency of AI-Generated Content. Final version published 2026-06-10; Commission Opinion of July 8, 2026, and AI Board assessment of July 9, 2026, on adequacy. Voluntary instrument for implementing Article 50 of the AI Act; adherence does not constitute conclusive evidence of compliance. https://digital-strategy.ec.europa.eu/en/news/commission-publishes-code-practice-marking-and-labelling-ai-generated-content; https://digital-strategy.ec.europa.eu/en/library/commission-opinion-assessment-code-practice-transparency-ai-generated-content.
• [SRC-S34-021] European Commission, General-Purpose AI Code of Practice. Final version published 2025-07-10. Voluntary instrument supporting compliance; GPAI obligations apply from 2025-08-02, with AI Office enforcement beginning on 2026-08-02 for new models and 2027-08-02 for existing models. Accessed: 2026-07-17. https://digital-strategy.ec.europa.eu/en/news/general-purpose-ai-code-practice-now-available.

Further Reading

These works expand on the book’s historical, political, and philosophical exploration. Unless they also appear in the Documentary Bibliography, they are not sources for its factual or legal claims.

• Bostrom, Nick. Superintelligence: Paths, Dangers, Strategies. Oxford University Press, 2014.
• Christian, Brian. The Alignment Problem: Machine Learning and Human Values. W. W. Norton & Company, 2020.
• Crawford, Kate. Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence. Yale University Press, 2021.
• Eubanks, Virginia. Automating Inequality: How High-Tech Tools Profile, Police, and Punish the Poor. St. Martin’s Press, 2018.
• Floridi, Luciano. The Fourth Revolution: How the Infosphere Is Reshaping Human Reality. Oxford University Press, 2014.
• O’Neil, Cathy. Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy. Crown, 2016.
• Russell, Stuart. Human Compatible: Artificial Intelligence and the Problem of Control. Viking, 2019.
• Zuboff, Shoshana. The Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power. PublicAffairs, 2019.