Roadmap for the Development and Application of Artificial Intelligence in Colombia

Scope definition for ethical and sustainable AI adoption

The roadmap addresses three fundamental pillars concerning artificial intelligence: adoption, ethics and sustainability.

Adoption of artificial intelligence

Technology adoption is a process of accepting, integrating and using new technologies to enhance quality of life and productivity, with barriers including access, affordability, generational gaps and knowledge gaps at social level. Government adoption of AI requires regulations protecting data and digital competencies, with Colombia having developed a database of 113 automated decision systems in the public sector documented across five components: basic information, data requirements, financing details, OECD classification and alignment with sustainable development goals, and information sources. Organizational adoption of AI involves integrating AI technologies into operations and production systems through machine learning, deep learning and natural language processing, though microenterprises with fewer than ten employees face barriers due to financial, human and organizational limitations and difficulty assessing cost-benefit ratios.

Ethics in artificial intelligence

AI ethics comprises moral principles guiding algorithm design and system outcomes to address bias, discrimination and ensure fairness. It is grounded in three Belmont Report principles: respect for persons through informed consent and protection of vulnerable groups; beneficence requiring algorithms to avoid harm and prevent bias amplification; and justice ensuring fair distribution of benefits and burdens across affected populations.

Sustainability in artificial intelligence

Defines three types of sustainability: environmental stewardship of natural resources, economically viable practices that are socially and environmentally responsible, and social cohesion and vital development. Requires that AI adoption in Colombia be sustainable to avoid contradiction with sustainable development goals, and demands transparent, responsible implementation decisions across all system layers to prevent negative social and economic impacts.

Why ethical and sustainable AI adoption: global situation and current challenges

Global AI adoption is accelerating with a 35% adoption index driven by implementation ease and cost reduction, but is concentrated in large enterprises with a 100% adoption gap increase since 2021 compared to small enterprises, facing barriers including limited AI skills at 34%, high cost at 29%, lack of development tools at 25%, project complexity at 24%, and data complexity at 24%. Ethical and sustainable AI adoption is imperative to address gaps, with most organizations lacking measures for trustworthy AI including bias reduction (74%), performance monitoring (68%), and explainability (61%), requiring UNESCO's human-centered approaches, OECD standards through the 2019 AI Principles, and national strategies such as those developed by seven Latin American and Caribbean governments including Colombia. The Inter-American Development Bank's fAIr LAC initiative promotes responsible AI adoption in social services through collaboration between public and private sectors, civil society and academia, addressing social challenges of equitable access, wealth distribution, and values promotion, though 85% of organizations lack bias-elimination protocols and nearly two-thirds report skills and training gaps despite 85% of consumers preferring companies with transparent AI model construction and use.

Commitment to ethical and sustainable transition

Companies increasingly view AI as a tool for advancing environmental, social and governance initiatives, with 64% of firms using AI already applying it to this end and 17% planning to do so. Global regulation of AI ethics remains fragmented, though cities, the EU through GDPR and the AI Act (expected to enter force in 2024), and regional initiatives like the fair LAC initiative have promoted digital rights-based policies and ethical frameworks; OECD principles adopted by six LAC countries require supporting inclusive growth and sustainable development, respecting human rights and democratic values, protecting privacy, data, non-discrimination and labour rights, and ensuring transparency and explainability throughout an AI system's lifecycle. AI systems must function without unreasonable safety risk under normal, foreseeable, incorrect and adverse conditions while maintaining traceability, risk management, privacy, digital security and bias mitigation, with AI actors held accountable for ensuring systems function well and respect ethical principles according to their specific duties, context and state of the art.

Baseline: Why ethical and sustainable AI adoption in Colombia?

Colombia faces the challenge of integrating AI ethically and sustainably into its society during rapid digital transformation, guided by foundations including CONPES 3975, Law 2069, Decree 1732, and UNESCO's recommendations on AI ethics emphasizing human rights, environmental protection, diversity, inclusion, proportionality, safety, equity, sustainability, privacy, transparency, and accountability. Colombia has developed 491 AI projects across public entities, established platforms for inclusive development and data governance, and achieved 9% AI adoption among companies primarily for administrative process optimization and operational efficiency, though adoption faces barriers including acquisition costs, lack of experience, and external obstacles such as supply chain disruptions and limited financial service access. Ethical and sustainable AI adoption requires multi-stakeholder governance involving governments, civil society, universities, private sector, and unions, with responsibility for AI implementation resting with those who program, train, and implement the technology rather than the technology itself, enabling precise regulation and advancement guided by international ethical principles.

Competitive advantages for ethical AI adoption

Colombia, ranking 6th on the Latin American Artificial Intelligence Index, possesses strategic advantages in human resources, academic output, and private sector participation for ethical and sustainable AI adoption, yet faces critical challenges including insufficient IT talent training (Latin America produces only 7% of needed IT talent), gaps in NBIC-related education alignment with productive sector needs, modest patent production, insufficient data processing infrastructure and connectivity, technological skills gaps, and gender disparities in AI skills. To strengthen its position as a Latin American AI leader, Colombia must foster five interconnected innovation environments (Ethics and Governance, Education and Research and Innovation, Data and Organizations, Innovative and Emerging Industries, and Privacy, Cybersecurity and Defence), enhance national AI strategies with specific regulatory frameworks, expand civil society and citizen participation, strengthen infrastructure and skills development programs with improved job transition policies, pursue regional collaboration and investment opportunities, and ensure AI development is guided by human dignity, human rights, gender equality, social justice, transparency in algorithmic functioning, and media literacy to address ethical challenges including discriminatory bias, digital divides, threats to cultural diversity and the environment.

Algorithmic transparency

Algorithmic transparency requires availability of information on algorithmic systems and automated decision systems through passive or reactive requests and active or proactive periodic disclosure, enabling accountability and exercise of rights including personal data protection and privacy. Colombia's Ethical Framework for AI addresses nine principles across data, algorithms, and practices: transparency and explanation requiring comprehensible information on data sources and system objectives with performance indicators; privacy requiring authorization and non-discriminatory design; human control requiring initial full oversight and human intervention on unwanted outcomes proportional to risk, with humans making critical decisions in sensitive sectors; accountability with collective liability unless single culpability demonstrated; non-discrimination prohibiting results threatening specific groups based on sex, race, religion, disability, age or sexual orientation; inclusion requiring diverse populations in design with representative data and gender-neutral systems; children's rights requiring superior interest priority, data use for positive impact only, comprehensible algorithm design, and minor participation in development; and social benefit requiring systems to generate cost reduction, productivity increases, public welfare, service provision, environmental protection or entertainment, with public sector systems pursuing other purposes not to be implemented. Implementation involves regulatory sandboxes under Decree 1732 of 2021 for innovation in regulated industries, with testing environments assessing compliance with existing regulations and regulatory changes, an intersectoral committee setting minimum requirements, and information sharing between sectors, plus organizational pre-development human rights impact assessments, data quality testing, accountability implementation, workplace transparency to employees and unions, employee understanding of decision-making logic and ability to address discrimination concerns, and security measures throughout the information lifecycle adapted to treatment risk and subject to continuous review.

Education, research and innovation

AI automation will change 44 percent of worker skills within five years, with 80 percent of the US workforce and 19 percent significantly affected by advanced language models, impacting women, minorities, indigenous people and low-income groups disproportionately. Inclusive education in AI through public systems, teacher training and curriculum incorporation from early stages is essential to address disparities and ensure equitable access to technology and learning opportunities. Colombia faces gaps between current and required human capital skills, requiring urgent changes to higher education and vocational training as global AI-driven transformations risk 47 percent of jobs; a new competencies-based educational model is proposed as more resilient to AI and developing critical thinking, ethics and emotional intelligence that AI lacks.

Data governance challenges

Data governance in artificial intelligence requires defining roles and responsibilities of data providers, consumers, service providers and regulatory authorities, with ethical frameworks and safeguards addressing transparency, accountability, explainability and risks to privacy and human rights. The Superintendencia de Industria y Comercio's data protection policy applies to all physical and digital databases and governs collection, storage, use, circulation and deletion of personal data under Colombian law 1581 of 2012, requiring adequate safeguards against abuse rather than prohibiting data processing. Artificial intelligence developers must comply with local personal data treatment rules, conduct privacy impact assessments, incorporate privacy, ethics and security from design and by default, demonstrate accountability, establish appropriate governance structures, implement measures for data treatment principles, respect data subjects' rights with effective exercise mechanisms, ensure data quality, use anonymization tools, and increase transparency with data subjects.

Industry 4.0 overview

Industry 4.0, involving fully automated production systems with digital technologies, the internet, cyber-physical systems, Internet of Things and Services, and Smart Factory, represents both strategic opportunity and risk for all sectors and requires new specialist competencies. Colombia ranks 62nd among 100 countries in production readiness and faces challenges in technology adoption, research investment, trade balance, economic-tax system complexity, cybersecurity risks, data breach vulnerabilities, and human capital barriers in science, technology, engineering and mathematics education. Colombia requires integrated efforts from government, private sector and academia addressing technological adoption, education, investment and infrastructure, and must strengthen human capital development, promote artificial intelligence adoption in enterprises, and increase economic resources dedicated to research and development.

Privacy, Cybersecurity and Defense

The aerospace and defence sectors valued at 29.67 billion dollars in 2023 and projected to reach 42.60 billion dollars by 2028 with annual growth of 7.5% can benefit from AI's real-time data processing for sensors, satellites and surveillance networks, and autonomous systems for reconnaissance and rescue missions, but autonomous decision-making raises accountability questions regarding responsibility in incidents and potential offensive operations, new vulnerabilities from malware and manipulation, and ethical challenges in delegating target selection to algorithms that cannot adequately discern complex situations or determine intentions. The United States, China and Russia compete as principal powers in this technological area while the United Nations seeks to limit lethal autonomous weapons through the Convention on Certain Conventional Weapons and the European Union's Artificial Intelligence Law emphasizes human responsibility and accountability; Colombia's Military Forces identified seven AI application areas including data collection and processing, surveillance and early detection systems, automation of routine tasks, improved precision and speed in decision-making, simulations and predictive models, predictive threat analysis, and detection and response to cyber threats. AI can help detect climate anomalies through oceanographic, hydrographic and meteorological data processing to develop early warning systems for nine climate-related risks including tropical cyclones, sea level rise, flooding, coastal erosion, forest fires, melting, droughts, landslides and increased jet stream intensity, and can monitor eight environmental risks including illicit crop cultivation, illegal mining, unauthorized logging, illegal fishing, chemical spills and ecosystem degradation in strategic areas using remote sensing and predictive models to track contaminants and generate environmental sensitivity maps.

Roadmap challenge and mission

The roadmap aims to ensure ethical and sustainable adoption of artificial intelligence aligned with OCDE and UNESCO principles while supporting Colombia's National Development Plan 2022-2026, based on the IBM Global AI Adoption Index 2023 surveying 8,584 IT professionals in 15 countries with a margin of error of plus or minus 1 percentage point and the Stanford AI Index tracking developments across 127 countries as of 2023. Colombia had 21 percent AI adoption relative to other Latin American countries in 2019, while globally 40 percent of organizations are exploring AI and 42 percent are implementing it; however, main barriers include limited skills and experience (33 percent), data complexity (25 percent), ethical concerns (23 percent), integration difficulties (22 percent), high costs (21 percent) and lack of development tools (21 percent), with fewer than half of implementing companies taking measures for trustworthy AI such as bias reduction, data provenance tracking or model explainability. In Colombia, 64 percent of survey participants trust that AI products have more benefits than drawbacks, and 58 percent report AI has profoundly changed their daily life in the past 3 to 5 years.

Challenge

Establishes the overarching challenge for the roadmap: to guarantee ethical and sustainable adoption of AI in Colombia.

Mission

By 2033, the strategy aims for half of Colombian organizations to have adopted artificial intelligence through responsible, ethical, and sustainable implementation. The strategy identifies five innovation focuses addressing low public understanding and trust in AI applications, insufficient competencies in AI development especially among women, absence of ethical frameworks in AI governance, low adoption rates in enterprises and government, and inadequate solutions to social, economic, and environmental challenges. For each focus, the strategy sets corresponding objectives including ensuring 100% of applications respect privacy and are transparent and fair, increasing women and underrepresented groups' participation by 50%, implementing ethical frameworks for 100% of AI data governance, achieving 50% adoption in enterprises and government sectors, and developing solutions addressing sustainable development goals.

Innovation focus and technology readiness level for AI

The strategy establishes five innovation focuses - Ethics and Governance, Education and Research and Innovation, Data and Organizations, Innovative and Emerging Industries, and Privacy, Cybersecurity and Defense - supported by four key factors: understanding and education, stakeholder engagement, institutional articulation, and AI democratization. Key implementation areas include risk evaluation and ethics assessments, privacy and personal data protection in predictive and prescriptive AI, intellectual property protection frameworks, scalable AI infrastructure and decentralized data access, citizen participation in AI policy through public consultation forums and citizen platforms, ethical auditing standards and education in AI ethics for professionals and the public, curriculum integration of AI across educational levels, specialized AI research and development centers with university-industry partnerships, and educational platforms for self-directed learning. The roadmap addresses data management practices including privacy standards and data audits, predictive analytics for business decision-making, personalized AI solutions across sectors, AI applications in supply chain management for demand forecasting and inventory optimization, advanced security solutions for threat detection, industrial automation through robotic systems and autonomous technologies, Internet of Things solutions for industrial interconnection, additive manufacturing integrated with AI, and industry digitalization through digital twins.

General objective

The roadmap aims to ensure that artificial intelligence is adopted in Colombia in ways that are both ethical and sustainable.

Specific objectives

Establishes five targets: ensuring all AI applications guarantee user privacy and are transparent/fair to build public confidence, increasing women and underrepresented groups in AI by 50% through training, implementing ethical frameworks so 100% of AI data governance follows ethical/legal principles, achieving 50% AI adoption among enterprises and government sectors, and developing AI solutions addressing social/economic/environmental issues aligned with national development goals.

Action plan

A total of 130 innovation routes undergo expert evaluation followed by public consultation to establish priorities and recommendations, identifying which routes most significantly advance the roadmap's objectives.

Routes for ethics and governance environment

Recommends exploring five high-impact topics in this area. Includes OECD recommendations for Colombia to develop national AI strategies in public sector, establish national data strategies, strengthen ethical frameworks, foster regional cooperation, support subnational initiatives, ensure long-term sustainability, and implement trustworthy AI governance with anticipated innovation techniques.

Routes for education, research and innovation

Proposes AI-based educational innovations including chatbots for student support, adaptive learning systems, intelligent tutoring platforms, collaborative learning with natural language processing, exploratory learning environments, automated writing assessment, language learning applications, and hybrid teaching models combining human educators with AI assistants.

Routes for data and organizations environment

Identifies AI capabilities for organizations including computer vision, natural language understanding, conversational interfaces, speech recognition, and customer segmentation. Recommends developing customized AI solutions for specific business needs and implementing AI in supply chain management for demand forecasting, inventory optimization, and logistics efficiency.

Routes for innovative and emerging industries

Proposes advanced automation in manufacturing using robotics and AI systems, creating digital twins of industrial processes for monitoring and optimization, and integrating AI to enhance industrial sustainability and energy efficiency through real-time monitoring and environmentally responsible production strategies.

Routes for privacy, cybersecurity and defense

Addresses cybersecurity threats including cyberattacks and AI-based adversarial attacks. Proposes AI surveillance systems for protecting natural resources and detecting illegal activities, intelligent waste management, environmental monitoring for air/water quality, AI-assisted urban planning, early warning systems for natural disasters, and smart monitoring with video analysis and facial recognition.

Monitoring and evaluation mechanisms

Establishes that the roadmap applies from 2024 to 2033 with semiannual review, beginning with a progress report in June 2024 and closing report on December 30, 2033. Indicates that monitoring and evaluation mechanisms will be implemented to ensure ethical and sustainable AI adoption.

Monitoring Mechanisms

Establishes methods to track progress on AI adoption goals through specific performance benchmarks and both numerical and qualitative measures covering privacy, openness, gender balance, and ethics. Requires regular surveys and feedback from Colombian residents and stakeholders to assess confidence levels in AI applications. Includes tracking participation in educational programs for underrepresented groups, particularly targeting a 50% increase in women's participation.

Evaluation Mechanisms

Prescribes periodic assessments of how widely AI is being adopted by organizations and the effectiveness of demonstration projects and training programs in building trust. Calls for evaluating how AI solutions address social, economic, and environmental challenges and whether they advance national development goals and sustainable development objectives. Includes regular reviews of research on ethics and governance frameworks, as well as monitoring workforce diversity in the AI field to meet the target of 50% participation by women and minorities.

References