commbox Artificial Intelligence & Algorithms

Artificial Intelligence & Algorithms

Filter

367

Sort by:

Direction:

Download (68)

Handbook on Data Protection and Privacy for Developers of Artificial Intelligence (AI) in India

Ajey Karthik, Aman Taneja, Sreenidhi Srinivasan, Vidushi Sinha

New Delhi: Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ); FAIR Forward – Artificial Intelligence for All (2026), 117 pp.

"This Handbook is designed as a practical guide for developers of AI systems, especially from early-stage startups, to navigate data protection obligations and ethical considerations, in a clear and actionable manner. Rather than serving as a comprehensive legal or technical manual, the Handbook co ... more

External Link (92)

The Essential Convergence: Global Compact on Extreme AI Risks

Sundeep Waslekar, Ilmas Futehally, Jayantika Kutty

Mumbai: Strategic Foresight Group (2026), 90 pp.

"Supply-Side Governance: Progress and Limitations: Most existing AI governance frameworks focus on the supply side of the technology ecosystem. These measures regulate developers, model providers, and infrastructure operators responsible for building advanced systems. Supply-side mechanisms include: ... pre-deployment safety testing and red-teaming; model evaluation and capability assessments; transparency and reporting obligations; incident response mechanisms; restrictions on certain high-risk uses. Several jurisdictions are experimenting with these tools in different combinations. China has implemented regulatory standards for generative AI systems; the European Union has adopted a risk based regulatory framework; Brazil and South Korea are developing legislative approaches; and the United States has introduced a mixture of executive measures and sectoral rules. These frameworks represent important progress. However, they face two structural limitations. First, supply-side regulation depends heavily on the cooperation of a small number of developer states. Second, the growing distribution of AI models, through open-weight releases, cloud APIs, and downstream applications, means that many systems reach global markets beyond the jurisdiction where they were originally developed. This reality limits the effectiveness of purely supply-side approaches. Demand-Side Governance: A New Dimension of AI Regulation: The paper introduces a second dimension of governance that has received far less attention: demand-side oversight. Demand-side governance shifts the locus of control from the point of creation to the point of deployment. Instead of relying solely on developer-state regulations, importing countries can establish safety requirements for any AI systems operating within their markets. These requirements may include risk assessments, certification standards, procurement rules, insurance obligations, and infrastructure safeguards. The central insight is that states controlling large consumer markets, digital infrastructure, financial systems, or public procurement channels possess significant leverage over global AI deployment, even if they do not develop frontier models themselves. This demand-side leverage can reshape incentives across the global AI ecosystem. Developers seeking access to major markets must comply with the safety conditions imposed by importing states. This approach represents a strategic shift in global AI governance: from a developer-centric model to a market-driven model of safety enforcement. Countries across different regions possess different forms of demand-side power. Brazil and South Africa represent large emerging markets; the United Arab Emirates plays a major role in AI infrastructure investment; South Korea combines technological capability with advanced regulatory institutions. When coordinated, these forms of leverage could help establish global safety norms. The Case for a Global Compact: Given the limitations of purely national approaches, the paper proposes the development of a Global Compact on Extreme AI Risks. The Compact would not aim to harmonise all aspects of AI regulation. Instead, it would focus narrowly on the most severe risk categories and establish shared mechanisms for prevention, monitoring, and response. Five institutional pillars form the backbone of the proposed Compact. The five pillars in the Global Compact should not be treated as isolated measures but designed as an interoperable system in which each pillar reinforces and informs the others through shared definitions, compatible standards, and coordinated implementation mechanisms. To address extreme risks effectively, the Compact should rest on a small set of guiding principles: a human-centric principle to ensure protection of humanity and human dignity as the primary objective; technological neutrality so that safeguards apply across methods and architectures rather than specific tools; interoperability to enable national and regional AI frameworks to exchange information, align risk classifications, and coordinate responses; and mutual cooperation to institutionalize cross-border collaboration in prevention, incident reporting, and crisis management. These can be complemented by principles of proportional risk governance, scientific integrity, and shared responsibility, creating a coherent normative base that allows diverse governance models to function together as a connected global safety architecture." (Executive summary, pages 5-6) more

External Link (196)

How Russia Is Reshaping Command and Control for AI-Enabled Warfare

Kateryna Bondar

Washington, DC: Center for Strategic and International Studies (CSIS) (2026), 20 pp.

"This paper examines how Russia is transforming its command and control (C2) architecture under wartime pressure, how these changes shape the country’s incremental move toward battlefield-required software solutions, and what lessons U.S. policymakers can learn from Russia’s experiences. Focusin ... g on both strategic ambitions and battlefield practice, the takeaways below summarize how automated C2 systems, unmanned platform management software, and emerging AI applications are being developed, adapted, and scaled within Russia’s military ecosystem. 1. Russia is no longer prioritizing the construction of a single, comprehensive automated C2 architecture comparable to Western joint concepts; instead, it is reallocating effort toward tactical, task-specific software, driven by battlefield necessity. Prolonged, high-intensity combat in Ukraine exposed the limits of centralized, system-wide C2 modernization and elevated the importance of accelerating the tactical kill chain. The emergence of systems such as the “Svod” Tactical Situational Awareness Complex and other integrated reconnaissance-strike tools reflects a pragmatic shift in which operational control of unmanned systems and real-time battlefield management now deliver greater military value than achieving end-to-end C2 integration. 2. Because unmanned systems now conduct up to 80 percent of Russian fire missions, the center of gravity in C2 innovation has shifted toward software that manages drones and integrates them with artillery and other fire units. Civilian engineers and volunteer developers have focused on closing this gap by building tools that provide situational awareness, automate fire correction, and link unmanned aircraft systems (UAS) operators directly to firing units. Russia’s “Glaz/Groza” software complex demonstrates this trend, functioning as a unified reconnaissance-strike workflow that converts drone footage into targeting data and compresses the time from detection to impact from hours to minutes. 3. The Russian military assesses its AI capabilities for visual and audio data processing as relatively mature, placing computer vision, sensor fusion, and signal analysis at technology readiness level (TRL) 6–9, while natural-language processing remains at an early, experimental stage, TRL 1–3. This disparity reflects a deliberate prioritization of AI applications that deliver immediate battlefield utility—such as target recognition, guidance, and autonomous terminal functions for unmanned systems—and where abundant data and combat validation are available. By contrast, text analysis AI, which underpins document processing and higher-level C2 decision support, remains constrained by immature architectures, limited certified software, and organizational barriers, slowing progress toward fully AI-enabled command workflows. 4. Within Russian C2 systems, AI is primarily envisioned as a support function rather than a replacement for human decisionmaking. Russian military doctrine assigns AI two core roles: enhancing the processing and interpretation of sensor data and providing predictive decision support through forecasting, scenario generation, and recommendations for commanders. Across strategic and tactical levels, AI is intended to augment situational awareness and analytical capacity, while formal authority and responsibility for decisions remain firmly with human commanders. 5. Russia began its C2 digitalization effort by building a dense layer of standards governing terminology, system architecture, hardware-software integration, and information management. This standardization drive, coupled with the transition to the domestically controlled Astra Linux operating system, reflects an attempt to create a unified technical foundation capable of supporting data integration, interoperability, and future AI insertion across the command hierarchy. While this framework provides structural coherence, it has not, on its own, resolved deeper institutional and methodological constraints that continue to limit system-wide C2 integration. 6. To enable AI-driven tactical software, the Russian military launched a systematic data collection effort in 2025 focused on unmanned operations and strike outcomes. The emerging data infrastructure aggregates UAS video feeds, operator telemetry, strike effects, and individual pilot performance metrics, each linked to unique personal identifiers. These datasets serve multiple functions simultaneously: operational analysis, training evaluation, and the creation of labeled data for AI model development, establishing a feedback loop that ties battlefield performance directly to software refinement. 7. Despite efforts to reduce dependence on foreign commercial technologies, Russia’s military AI development remains heavily reliant on open-weight models and civilian software ecosystems. The transition from tools such as AlpineQuest and Discord toward domestic alternatives like ZOV Maps and Astra-based platforms reflects a push for security and sovereignty at the application layer. At the same time, Russian developers actively adapt open-weight and commercially available AI models, including Mistral, Qwen, LLaMA, YOLO, and related architectures, for military use, embedding them into on-premise, tightly controlled environments. This hybrid approach allows Russia to mitigate sanctions and accelerate AI adoption without building foundational models from scratch." (Executive summary) more

Download (83)

Companion AI: Risiken von sykophantischen und suchtfördernden Designs von KI-Systemen, ihre rechtliche Erfassung und Handlungsempfehlungen
Deep Insights

Ramak Molavi-Vasse’i

Berlin: Zentrum für Digitalrechte und Demokratie (2026), 82 pp.

"Diese Studie untersucht Risiken von KI-Systemen, die im fortlaufenden Dialog mit Nutzern Emotionen erkennen, persönliche Bedürfnisse adressieren und darauf reagieren. Chatbots mit dieser Funktionalität werden als Companion-AI bezeichnet. Sie sind hochgradig personalisiert und treten Nutzern als ... soziales Gegenüber entgegen, das freundschaftliche, romantische oder sexuelle Nähe simuliert. Menschen können zu solchen Systemen eine emotionale Bindung aufbauen. Der Fokus liegt nicht nur auf spezifischen Companion-Anwendungen wie Replika oder Character.AI, sondern auch auf Universalmodellen wie ChatGPT, Claude, Gemini, Grok oder Meta-AI. Diese Systeme werden zunehmend für persönliche, emotionale und beratende Gespräche genutzt. Die Untersuchung kommt zu dem Ergebnis, dass Companion-AI mehrere eng miteinander verbundene Risikodimensionen erzeugt. Psychische und physische Gesundheit: Companion-AI trifft auf eine Gesellschaft, in der Einsamkeit und psychische Belastungen zunehmen, während professionelle Versorgungsangebote knapp sind. Die möglichen Risiken sind klinisch belegt. Die Nutzung kann psychische Belastungen verursachen oder verstärken und in Einzelfällen gravierende Gesundheitsfolgen auslösen. Dokumentiert sind die Verschlimmerung psychotischer Zustände, die Verstärkung depressiver Muster und Angststörungen, suchtartige Bindungen mit Entzugssymptomen sowie die Erosion sozialer Kompetenzen, etwa eine messbar reduzierte Konfliktfähigkeit nach längerer Interaktion mit Companion-AI. Öffentlich bekannt gewordene Vorfälle sind in der Vorfall-Datenbank der Studie dokumentiert. Eingriff in die Privatsphäre: Companion-AI animieren Nutzer kontinuierlich zur Selbstoffenbarung und greifen damit in sensible Gedanken und die intime Gefühlssphäre der Nutzer ein. Zugleich ermöglicht die fortlaufende Interaktion eine zunehmend verdichtete Profilbildung. Entscheidungsautonomie und demokratische Meinungsbildung: Companion-AI können sowohl die Qualität von Informationen als auch die Entscheidungsautonomie der Bürger beeinträchtigen. Die Mechanismen, die Nähe und Vertrauen erzeugen, beeinflussen zugleich die Generierung, Aufnahme und Gewichtung von Informationen. Sykophanz, also die unkritische Bestätigung von Nutzeransichten, beeinträchtigt Genauigkeit und Verlässlichkeit der Antworten messbar. Sprachmodelle werden zunehmend als primäre Quelle der Informationssuche genutzt. Wenn dieselben Systeme Informationen erzeugen, beschaffen und präsentieren, fallen Selektion und Aufbereitung in einer Hand zusammen. Werbe- und interessengeleitete Einflussnahme greift dann nicht mehr nur in einzelne Kaufentscheidungen ein, sondern in die Voraussetzungen öffentlicher Meinungs- und demokratischer Willensbildung." (Executive summary, Seite 5) more

External Link (70)

We Have an Artificial Intelligence Definition Problem in ICT4D

Wayan Vota

ICTworks (2026)

"A new study in the International Journal of Research and Innovation in Social Science surveyed 200 women micro-entrepreneurs in Bengaluru and reported that 32.5% of them are “actively using at least one AI-enabled tool” in their businesses. The tools, according to the paper, include KhataBook ( ... more

Download (47)

Justicia algorítmica en América Latina y el Caribe: Investigación, políticas públicas y ética de la inteligencia artificial

Consejo Latinoamericano de Ciencias Sociales (CLACSO); CIESPAL; UNESCO (2026), 105 pp.

"El volumen reúne los policy briefs elaborados en el marco de la convocatoria Justicia Algorítmica. Una inteligencia artificial ética para América Latina y el Caribe, lanzada conjuntamente por UNESCO Ecuador y el Consejo Latinoamericano de Ciencias Sociales (CLACSO), en el marco de una colaborac ... ión intensa, sostenida y productiva que ambas instituciones han consolidado en los últimos años. A ella se sumó El Centro Internacional de Estudios Superiores de Comunicación para América Latina, CIESPAL, responsable de la edición del libro en alianza con CLACSO y UNESCO, reafirmando la convicción de que la cooperación interinstitucional regional constituye un instrumento fundamental para pensar críticamente la tecnología, fortalecer la cultura democrática y promover políticas públicas capaces de atender a los grupos más vulnerables. Los trabajos seleccionados y aquí publicados son el resultado de un proceso riguroso, sistemático y situado: se recibieron 35 postulaciones, evaluadas en primera instancia por un Comité Internacional integrado por nueve especialistas provenientes de Ecuador, Argentina, Brasil y Francia. Posteriormente, UNESCO y CLACSO realizaron la revisión final atendiendo a la calidad técnica y científica, la coherencia metodológica, la consistencia del plan de trabajo y la pertinencia de los temas en relación con la Recomendación sobre la Ética de la IA y el equilibrio regional. El proceso concluyó con la selección de cinco propuestas, varias de ellas con participación de investigadoras e investigadores con trabajo en y sobre Ecuador. Los policy briefs que conforman este libro abordan problemas urgentes de la agenda regional sobre IA. Uno de los núcleos críticos es la transformación del trabajo y la desigualdad digital, un fenómeno particularmente agudo en América Latina y el Caribe, donde históricas brechas educativas, económicas y territoriales se ven amplificadas por una automatización que concentra oportunidades en sectores altamente especializados y desplaza tareas tradicionalmente realizadas por trabajadores menos calificados. La región enfrenta una doble asimetría: la del acceso desigual a tecnologías avanzadas y la de las capacidades necesarias para utilizarlas de modo significativo, reproduciendo patrones de exclusión que atraviesan género, etnia y condición socioeconómica. Otro foco central es la privacidad y la vigilancia algorítmica, ámbitos donde los riesgos son crecientes. La implementación de sistemas opacos —las conocidas “cajas negras” algorítmicas— que procesan datos personales a gran escala ha generado nuevas formas de discriminación, como el racismo algorítmico, sesgos de género, o modelos predictivos que penalizan a grupos históricamente vulnerados. Los casos de identificación errónea en sistemas de reconocimiento facial o la clasificación automatizada de supuestos “riesgos sociales” muestran que, sin marcos éticos y normativos sólidos, la IA puede reforzar desigualdades en vez de corregirlas." (Introducción, páginas 8-9) more

External Link (185)

Promoting and advancing human rights in global AI ecosystems: The need for a comprehensive framework under international law

Alexander Kriebitz, Caitlin C. Corrigan (eds.)

Globethics Foundation; Responsible AI Network – Africa (RAIN-Africa) (2026), 33 pp.

"In response to far-reaching societal impacts of AI, policymakers are increasingly adopting a human rights lens for AI governance. However, the regulatory landscape remains fragmented and characterized by Western led initiatives as well as inconsistent enforcement. To address these limitations, a un ... ified and binding international framework is urgently needed. The proposed Convention on AI, Data, and Human Rights (drafted as the “Munich Convention on AI, Data, and Human Rights”) is an initiative toward such a framework. Developed with contributions from over 50 global experts, it proposes an international convention to align AI governance with human rights principles. This whitepaper explores the underlying rationale behind the convention, emphasizing its alignment with the norms and principles articulated in the Universal Declaration of Human Rights and the UN Charter. It examines the relevance of safeguards against AI’s misuse in facilitating human rights violations, with particular focus on protecting underrepresented or marginalized groups, including indigenous peoples, and addressing collective rights. It furthermore points to the necessity to codify rights that empower individuals to opt out, be forgotten, seek explanations, and access remedies to maintain adequate human rights standards in the context of AI. The urgency to act arises from the escalating impact of AI and the rapid pace of political and technological advancements. The UN Human Rights Council (UNHRC) is particularly well poised to play a leading role in initiating global discussions on a binding convention for AI governance, grounded in human rights principles. Drawing on its established mandate to uphold human rights globally, the UNHRC has a record of successfully embedding human rights principles into international frameworks, as demonstrated by its endorsement of the UN Guiding Principles on Business and Human Rights and the recognition of the right to a clean, healthy and sustainable environment. Through dedicated action around promoting a convention on AI and human rights, the UNHRC can foster international consensus and proactively shape the future of AI in a manner that prioritizes the values and principles enshrined in the key documents of international human rights law." (Executive summary) more

Google Scholar (41)

WorldCat (37)

How to Talk to AI (And How Not To)

Jamie Bartlett

London: Ebury Publishing (2026), 272 pp.

"[...] Most people still don't really understand how AI works, how to make the most of it – or what the dangers are. As some people use it to turbo-charge their productivity at work, others are falling into dangerous conspiracies, delusions and psychosis. In How to Talk to AI, award-winning techno ... more

External Link (186)

Why Silicon Valley is turning to the Catholic Church: Priests and theologians want to shape the future of AI. Big Tech is listening

Elias Wachtel

The Atlantic, April 25 (2026)

"Silicon Valley and the Vatican may “seem like strange bedfellows”—but they each have something to gain from the other, @EliasWachtel argues. He examines why Big Tech is listening to priests and theologians who want to shape the future of AI." (Publisher description)

Download (80)

Handbook on Anthropology and Artificial Intelligence
Deep Insights

Sahana Udupa, Peter Hervik (eds.)

Cheltenham (UK); Northampton, MA (US): Edward Elgar Publishing (2026), xix, 495 pp.

"This timely Handbook provides a groundbreaking collection of anthropological research on artificial intelligence (AI), examining how anthropology can help to comprehend and critique its development. Evaluating the limits, hopes and fears of AI, leading experts explore its influence as a sociocultur ... more

Download (77)

Environmental Cost of AI's Energy Use: Carbon, Water and Land Footprints

Miriam Aczel, Sanaz Chamanara, Mir Matin, Aria Farsi, Tshilidzi Marwala, Kaveh Madani

Richmond Hill, Ontario (CA): United Nations University Institute for Water, Environment and Health (UNU-INWEH) (2026), 55 pp.

"[...] The study reveals some striking numbers. In 2025, data centers—the physical backbone of AI—consumed an estimated 448 TWh of electricity. If data centers were a country, that level of electricity use would rank it 11th globally. On current trajectories, data center electricity demand could ... roughly double to 945 TWh by 2030, nearly triple the combined annual electricity use of Pakistan, Bangladesh, and Nigeria, together home to more than 650 million people. Producing that much electricity would have a carbon footprint of 399 million tonnes CO2e, requiring 6.7 billion trees grown over 10 years to offset—roughly twice the number of trees in the United Kingdom. The associated water footprint of 9.3 trillion liters would be equivalent to the annual domestic water needs of all 1.3 billion residents of Sub-Saharan Africa. The land footprint associated with this electricity would exceed 14,500 km², nearly 10 times the size of Mexico City. AI is now one of the most significant drivers of that data center growth. In 2025, AI workloads alone accounted for around 20% of total data center electricity use, and if that share rises to 40% by 2030 as projected, its electricity demand could reach roughly 378 TWh—enough to meet the residential electricity needs of the entire population of Sub-Saharan Africa for over 2 years. AI’s environmental impacts are shaped not only by data center growth and electricity supply mixes, but also by the escalating cost of building ever-larger models. For example, GPT-3 training consumed an estimated 1.3 GWh of electricity over 34 days, while GPT-4 is estimated to have consumed 50 to 70 GWh over 100 days, roughly 40 to 55 times GPT-3. Yet training is only part of the picture as AI’s operational footprint is increasingly driven by inference. Once models are deployed, billions of everyday interactions account for the bulk of energy use, with inference estimated at roughly 80–90% of total energy consumption. [...] The report identifies priority implications for key stakeholders. Governments must integrate AI infrastructure into energy system planning, carbon accounting, water governance and land use permitting. This requires standardized environmental footprint reporting so impacts can be verified and compared across providers and jurisdictions. Industry and AI developers should treat model selection, default outputs, and routing decisions as footprint determinants, while also improving efficiency-by-design. Users and deploying organizations also shape impacts through volume, frequency, and modality choices. They should adopt “fit-for-purpose” use: selecting the lightest model and lowestenergy format that meets the task and limiting high-cost features when not needed. Data center operators and utilities should recognize siting and procurement as environmental footprint decisions, apply environmental impact screening and cumulative impact assessment. They should implement transparent mitigation and community safeguards where expansion occurs. Investors and financiers should treat electricity, carbon, water, and land footprints as material risks for AI infrastructure portfolios and use comparable footprint metrics in due diligence." (Executive summary, pages 12-13) more

External Link (131)

Principled caution: How AI is used in public-interest media in Asia

James Gomez

Bonn: Deutsche Welle DW Akademie (2026), 52 pp.

"In conclusion, this research has provided a systematic understanding of AI adoption within public-interest media across Afghanistan, Bangladesh, Bhutan, Cambodia, Indonesia, Myanmar, Pakistan and Uzbekistan. By identifying concrete areas for strategic intervention, this research seeks to strengthen ... journalistic quality and prepare the sector for an AI-integrated future. The findings reveal that the AI adoption is currently taking place within public-interest media, but it is one of “principled caution” where adoption is careful, selective and highly varied. This evolution unfolds within a structurally precarious media environment where AI regulation is often absent or actively restrictive. While adoption remains cautious to protect journalistic credibility from risks such as hallucinations and biases, newsrooms are increasingly embracing AI as a powerful toolset to alleviate severe capacity and resource constraints. Currently, organizations primarily utilize AI for assistive, routine tasks such as translation, transcription, summarization, and early-stage content ideation and the final language editing stage rather than for broader organizational transformation. The research highlights deep capacity deficits in the governance framework, institutional training, and leadership. These gaps often prevent individual AI literacy from translating into consistent, organization-wide policy formulation and enforcement. This challenge is further exacerbated by the funding instability and resource constraints that characterize the region’s independent media, restricting the long-term investment necessary for ethical and responsible AI adoption. Despite these legal, political, technical, capacity, and operational constraints, localized and innovative AI workflows are emerging. While AI has not yet generated entirely new, revenue-generating business models, strategic investments in customized AI infrastructures, such as in-house dashboards and media monitoring systems, present a critical pathway toward long-term operational resilience. To support this trajectory, the report recommends coordinated action among media organizations, governments, technology companies, donors and media development funders to invest in long-term AI capacity, governance frameworks, and secure technological infrastructure." (Conclusion, page 39) more

External Link (128)

AI in the media: Kenya and South Africa

Karen Allen, Herman Wasserman, Nande Mbekela

Bonn; Stellenbosch (ZA): Deutsche Welle DW Akademie; Centre for Information Integrity in Africa (CINIA) (2026), 27 pp.

"This study shows that AI remains largely an underexplored, and often misunderstood, topic for journalism in South Africa and Kenya. Stories can be framed in overly positive terms, such as the potential AI holds for development and social change, or negative terms, by highlighting the extractive nat ... ure of big tech firms or the potential of AI to foster disinformation. Either way, these reports frequently lack a more nuanced understanding of the impacts of AI in African contexts. Fear of the impact of AI on the media workforce may also explain why reporters say they often face resistance by editors when seeking to get AI and tech stories commissioned. AI articles tend to be siloed and reactive, framed as business, security or technology stories. Seldom is AI presented in terms of its multidimensional, layered impact on society. Furthermore, the geopolitical aspect of AI rarely gets interrogated in mainstream media. Journalists interviewed for the study admitted that many of their stories were reactive, and identified areas where synergies could be harnessed from other parts of the world. On issues of AI and social media regulation, for example, where precedents have been set by countries such as Australia, South African and Kenyan media could use such precedents to develop locally focused stories. AI reporting often adopts a framing promoted by technology vendors who derive interest from presenting a one-dimensional, optimistic view of AI. The unintended consequences often go unreported unless supported by an investigation, such as the CCTV camera surveillance reportage by Swart referenced above. Furthermore, the topic is often presented through a Global North lens, with little contextualisation or interrogation of these issues from an African point of view amid resource and access constraints. This occurs against a background of under-developed policy frameworks for AI on the continent generally, and in newsrooms in particular. Consequently the playing field can be tilted in favour of AI and tech vendors who fill the vacuum with hyperbole and hype about what the technology is capable of delivering. At the same time, more nuanced and specialised reporting on AI is constrained by the familiar obstacles of journalism, such as lack of resources, over-stretched newsrooms and story competition. To enhance coverage of AI in Africa, a multifaceted strategy is needed. While training and AI literacy programmes can deepen the knowledge needed by African journalists to investigate the topic more comprehensively, increased investment in newsrooms can provide the structural support to provide more context-specific coverage of AI on the continent. Crucially, training and education of African journalists should be done from within the African context, as solutions cannot be imposed from newsroom contexts in the Global North." (Conclusions, page 21) more

External Link (144)

Google Scholar (173)

WorldCat (158)

Communicating Sustainable Development Goals in the Age of Artificial Intelligence

Muhammad Jameel Yusha'u

Palgrave Macmillan (2026), xlvii, 523 pp.

"This book is the first scholarly work dedicated to examining the complex relationship between artificial intelligence and communicating the UN Sustainable Development Goals. Moving beyond techno-optimist narratives, the book advances a critical theoretical intervention: the AI Sustainability Exclus ... more

Google Scholar (24)

WorldCat (25)

Hallo, wie fühlst du dich heute? Wie künstliche Intelligenz unsere Beziehungen verändert

Christoph Koch

Zürich: Kein & Aber (2026), 272 pp.

"Christoph Koch hat ein Jahr lang mit Menschen gesprochen, die mit KI leben — mit denen, die sich verlieben, denen, die trauern, denen, die Chatbots als Therapie nutzen. Er hat mit KI-Forscherinnen und Psychologen geredet. Mit Anwälten und Geschädigten. Und mit den Maschinen selbst. Herausgekomm ... more

Download (12)

Taking a Look at the most common AI myths & debunking them the Superrr way!

Berlin: SUPERRR (2026), 22 pp.

"How do we want to use technology, and when do we hit the brakes? As part of our work in 2026, we held 14 public-facing “AI mythbusting” sessions at libraries across Berlin, Germany. We were astonished by the strong public response and clear need for judgement-free spaces to talk about AI. Colle ... more

Download (41)

Uso crítico y ético de la Inteligencia Artificial en la educación
Deep Insights

Santiago García Gago, Inés Binder

São Paulo: Campaña Latinoamericana por el Derecho a la Educación (CLADE); Educación en Voz Alta (2026), 150 pp.

"Esta guía sistematiza y resume los contenidos del curso virtual “Uso crítico y ético de la Inteligencia Artificial en la educación”, celebrado en tre los meses de octubre y diciembre de 2025 [...] El curso se planificó como una introducción conceptual a los distintos ámbitos que abarca ... la Inteligencia Artificial, entendida como un sistema sociotécnico complejo y multidimensional. La idea es que estos conceptos apoyen a docentes y educadores en el diseño de estrategias que incorporen la IA crítica y éticamente en los procesos educativos. Por lo tanto, no es manual técnico de aplicaciones ni tampoco uno de estrategias pedagógicas. Se concibió como un compendio de fundamentos teórico-prácticos que sustenten el diseño de las estrategias para incluir aplicaciones de IA en las aulas. Para ello, en el capítulo primero, se desmitifica el desarrollo tecnológico desde la sociología y la economía política preguntándonos quiénes son los dueños de la IA y qué intereses económicos y geopolíticos dirigen su destino. En el capítulo dos se abordan los conceptos técnicos, las definiciones y tipologías que permiten abrir la caja negra de la Inteligencia Artificial y entender cómo funciona. También se enumeran algunas aplicaciones, tanto las privativas como modelos open source que permiten experimentar en el ámbito educativo o usar con mayores niveles de privacidad estas herramientas. El capítulo tres, alejándose de la lógica instrumental que analiza las tecnologías desde los usos “buenos” y “malos”, problematiza las relaciones de poder que configuran la IA y los desafíos a los que nos enfrenta clasificados en siete dimensiones: técnicas, éticas, ecológicas, de género, políticas, laborales y de seguridad y privacidad. Y cierra con una breve exposición de los esfuerzos legislativos, tanto a nivel internacional como regional y nacional, para regular la Inteligencia Artificial. En el capítulo cuarto, se exponen diversos aspectos cognitivos y del El capítulo quinto, complementando el anterior, aborda estrategias concretas para la inclusión ética de la IA en los procesos de aprendizaje desde tres abordajes: con pedagogías centradas en la persona, integrales e interdisciplinares y basado en la tecnología y no en la aplicación." (Introducción, páginas 9-11) more

Download (81)

Step by Step: How SAFE AI contributes to trustworthy AI

CDAC Network; Alan Turing Institute; Humanitarian AI Advisory (2026), 8 pp.

"Based on the National Institute of Standards and Technology (NIST) AI Risk Management Framework, the seven characteristics of trustworthy AI are: 1. Valid and Reliable: The AI system should perform as intended, delivering accurate results consistently across different situations. 2. Safe: The sy ... more

External Link (180)

AlgorithmWatch’s guidelines to use generative AI responsibly

Oliver Marsh, Michele Loi

Berlin: Algorithm Watch (2026)

"Whether you use ChatGPT, Claude or Gemini, Copilot or Perplexity – generative AI poses massive problems: many results are inaccurate and politically problematic, the systems’ energy and water consumption is enormous. At the same time, they have become an integral part of everyday life. Algorith ... more

Download (49)

Indic Voice Technologies for an Inclusive Digital India: Toolkit for Developers

Nihar Desai, Sujith P., Harleen Kaur, Dona Mathew, Jyotsna Jayaram, Padmavathi Prasad, Thomas Vallianeth

Bhashini; Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ); FAIR Forward – Artificial Intelligence for All (2026), 70 pp.

"This toolkit sets out a layered, lifecycle-oriented approach to building inclusive and robust speech artificial intelligence (AI) systems. It brings together strategies for diverse and representative data collection, linguistically informed model training, rigorous quality control, and deployment o ... more

Companion AI: Risiken von sykophantischen und suchtfördernden Designs von KI-Systemen, ihre rechtliche Erfassung und Handlungsempfehlungen Deep Insights

Handbook on Anthropology and Artificial Intelligence Deep Insights

Uso crítico y ético de la Inteligencia Artificial en la educación Deep Insights

Companion AI: Risiken von sykophantischen und suchtfördernden Designs von KI-Systemen, ihre rechtliche Erfassung und Handlungsempfehlungen
Deep Insights

Handbook on Anthropology and Artificial Intelligence
Deep Insights

Uso crítico y ético de la Inteligencia Artificial en la educación
Deep Insights