Human‑Centric AI and National Capability Roadmap from ICAN 2026

The conference opened with a call to shift AI development from pure technology acquisition to building human agency, judgment, and ethical control. The speaker framed AI as a capacity rather than a product, emphasizing three pillars—robust infrastructure, comprehensive education, and workforce readiness. “Our greatest global challenge is no longer access to technology. It is the cultivation of human capability,” was declared, setting the tone for the sessions that followed.

Institutional Capacity & Universities

Universities were positioned as the primary platforms for research, talent development, and technology translation. Rather than confining AI to computer‑science departments, the agenda urged integration across all disciplines, including the social sciences. Assessment models must be redesigned to reward critical thinking and problem‑solving instead of mere tool dependency. The message was clear: AI leadership is defined by developing minds, not just importing models.

National Data & AI Engine

A data‑centric engine was proposed to standardize transformation pipelines and capture full data lineage. Because 80 % of AI work involves data preparation, a unified approach prevents “reinventing the wheel.” The Data Commons model illustrates how knowledge graphs can link disparate sources, while hybrid compute—centralized training paired with edge inference—offers flexibility in storage and scaling. Sovereign AI tiers were outlined, ranging from commercial API use (Level 0) to training home‑grown models such as the “Alam” LLM (Level 3).

Governance, Ethics, and Human Agency

Effective oversight requires a single entity with “strong teeth” to enforce mandates on liability, data ownership, and safety checks. While data trust appears elegant on paper, practical implementation is messy, demanding clear accountability. Cognitive offloading—delegating judgment to AI—has already caused a reported 65 % drop in critical thinking in some studies. Consequently, human‑in‑the‑loop remains mandatory for high‑stakes domains like healthcare to combat automation bias and preserve strategic agency.

Human Capability Development

A three‑pillar approach to human capability was introduced: AI recognition, AI usage, and AI creation. The “Human Code”—connection, reasoning, and creativity—defines the skills needed to coexist with abundant intelligence. The speaker warned that when intelligence is abundant, humans are not obsolete, but agency becomes scarce. Reskilling programs such as the “Jahizia” readiness test, which reached 36 000 students in 2025, aim to embed these capabilities across the population.

Strategic Execution

Many AI projects fail because they are treated as isolated technology pilots rather than end‑to‑end process reinventions. To avoid “pilot graveyards,” the focus must shift to measurable business outcomes. The “translator” mechanism bridges public, private, and academic sectors, aligning timelines and expectations. Regional success also depends on easing cross‑border movement for startups and establishing unified data‑sharing agreements. Europe’s lag, the speaker noted, stems from over‑regulation and a weak startup ecosystem.

AI as a Research Collaborator

AI was described as “augmented intelligence” that accelerates literature reviews and complex system analysis. Researchers act as “verifiers,” retaining responsibility for scientific outcomes while leveraging AI‑generated insights. Knowledge distillation and model quantization enable efficient, domain‑specific models, but the ultimate driver of breakthroughs remains human curiosity, which AI cannot replicate. Responsible AI must become part of delivery, not a compliance afterthought.

Mechanisms in Practice

• Cognitive Offloading reduces critical thinking when users rely excessively on AI.
• Knowledge Distillation creates smaller, efficient models by emulating larger ones.
• Model Quantization cuts precision to improve computational efficiency.
• Federated Learning preserves privacy by sharing only aggregated updates.
• Data Lineage ensures reproducibility by documenting every transformation step.
• Agentic AI Oversight imposes strict action boundaries and rollback protocols for autonomous systems.

These mechanisms underpin the broader strategy of building a sovereign, human‑centric AI ecosystem.