AI literacy – Artificial intelligence has entered education faster than education systems have learned how to govern it.
That is no longer a prediction. UNESCO reports that 92% of higher-education professionals already use AI tools in their work, while only 23.6% say they feel very confident using them.[1] The distance between those two numbers may prove more consequential than the rate of adoption itself.
It tells us that access is accelerating faster than capability.
For ICARUS, this is becoming one of the defining education questions of the AI era. Giving teachers and learners access to increasingly powerful systems is important, but access alone cannot produce responsible use, sound judgment or educational value. The next challenge for AI literacy in education is therefore deeper: people need the capability to question AI, verify it, understand its limitations and remain accountable for the decisions they make with it.
We call that AI agency.
AI adoption has moved ahead of confidence
The first phase of generative AI in education focused heavily on access. Institutions debated which tools students could use, whether teachers should incorporate them, how assessment might change and whether AI-generated work should be restricted.
Those questions remain relevant. However, the technology has moved much faster than the debate.
Teachers now use AI for lesson planning, research, writing, summarization and content development. Students use it to explain concepts, generate ideas, translate material and support assignments. Universities are incorporating AI into research and administration.
As adoption becomes normal, the central question changes.
Knowing how to enter a prompt is a very low threshold for AI literacy. An educator also needs to recognize when an answer lacks evidence, when a source has been fabricated, when an automated recommendation embeds bias, or when using AI may undermine rather than improve the learning process.
The real educational task is no longer simply teaching people how to use artificial intelligence. It is developing the judgment required to use it well.
From AI literacy to AI agency
AI literacy is often described as knowledge about artificial intelligence: how it works, what it can do and what risks accompany it.
That foundation matters. Yet education must go further.
AI agency in education means that teachers and learners retain the ability to understand, question, verify and ultimately decide.
A person with AI agency does not surrender judgment simply because a system produces a convincing answer. They ask where information came from, whether evidence supports it, what assumptions shaped the result and what human consequences may follow from using it.
This distinction becomes particularly important as AI systems become more fluent and persuasive. The greatest educational risk may eventually come from outputs that sound sufficiently credible that people stop examining them carefully.
For that reason, critical thinking cannot become a casualty of AI adoption. It must become one of its central competencies.
UNESCO’s own AI Competency Framework for Teachers points in this direction. It defines 15 competencies across five dimensions: a human-centred mindset, ethics of AI, AI foundations and applications, AI pedagogy, and AI for professional learning.[2]
The message is significant. Technical proficiency represents only part of AI readiness.
What should an AI-ready teacher actually be able to do?
The phrase “AI-ready education system” risks becoming meaningless unless we can describe the human capability behind it.
For ICARUS, an AI-ready teacher should be able to understand the basic logic and limitations of the systems they use. They should know how to interrogate an output rather than accept it because it sounds authoritative. They should be able to verify important claims against credible evidence and recognize situations where an automated response deserves additional scrutiny.
Moreover, teachers need the confidence to decide when AI strengthens pedagogy and when it weakens it.
Sometimes AI may help a teacher translate learning materials, locate information more efficiently or adapt an explanation to different levels of understanding. In other situations, introducing AI may remove the intellectual effort through which learning actually occurs.
That is a pedagogical judgment, not a technical one.
The same principle applies to privacy, fairness, assessment and intellectual responsibility. Ultimately, an educator must remain responsible for the educational decision, even when technology contributes to it.
AI literacy therefore has to develop into professional capability.
An ICARUS framework for AI agency
We believe that capability can be organized around five connected actions:
Understand → Question → Verify → Decide → Apply
Understanding means knowing enough about AI to recognize both its usefulness and its limitations. Questioning means approaching outputs with intellectual independence rather than technological deference. Verification requires checking important claims, evidence and sources. Decision means preserving human accountability over whether and how AI enters an educational process. Application means using AI purposefully where it creates genuine educational value.
Together, these actions describe something more demanding than tool proficiency.
They describe agency.
The distinction matters because an educator may become highly proficient with an AI interface while remaining poorly prepared to evaluate what the system produces. Conversely, a teacher does not need to become an AI engineer to exercise informed judgment.
The objective should therefore be practical competence rather than technical mystification.
Teachers need enough understanding to remain intellectually and professionally in control.
Measuring AI agency instead of assuming it
The next step is measurement.
At ICARUS, we believe AI literacy initiatives should demonstrate whether participants actually develop better judgment. A programme should therefore examine capability before and after learning rather than report only attendance or completion.
For example, educators could receive realistic AI-generated outputs containing factual errors, unsupported claims, fabricated references, subtle bias or inappropriate pedagogical recommendations. Assessment could then measure whether participants recognize the problem, explain why it matters, identify a verification route and make an appropriate decision.
Confidence also matters, although confidence alone cannot serve as proof of competence. A strong evaluation model would compare self-reported confidence with demonstrated ability.
Over time, institutions could also examine whether educators introduce verification routines into their teaching, whether they make more deliberate decisions about AI use, and whether learners become more capable of challenging machine-generated information themselves.
This creates a pathway from AI awareness toward measurable AI agency.
Why this matters for SDG 4
AI agency also belongs inside the wider conversation about SDG 4: inclusive and equitable quality education and lifelong learning for all.
Artificial intelligence may expand access to personalized support, translation and educational resources. At the same time, unequal capability can create a new layer of educational inequality.
Two learners may have access to the same AI system while receiving very different benefits from it. One may know how to interrogate, refine and verify what the system produces. Another may accept the first answer as knowledge.
The digital divide is therefore changing.
Connectivity still matters. Access still matters. Language still matters. Increasingly, however, the ability to exercise judgment within digital systems matters as well.
That is why responsible AI education should become part of the quality dimension of SDG 4. Access to technology without the capability to use it intelligently does not constitute educational inclusion.
From the UNESCO Knowledge Hub to the next question
This thinking also grows naturally from work that UNESCO has already published about ICARUS.
The UNESCO SDG 4 Knowledge Hub features the ICARUS practice “AI-enabled multilingual learning infrastructure for inclusive lifelong learning.” The case study addresses a related problem: digital access does not automatically translate into understanding, participation or competence.[3]
The practice uses AI-enabled video indexing, transcription, translation, searchable knowledge repositories, adaptive discovery and competence-oriented learning pathways. Its underlying principle is straightforward: technology becomes educationally useful when it reduces barriers and strengthens human learning.
AI agency takes that principle into the next phase.
The question is no longer only whether learners can find and understand knowledge. It is increasingly whether they can exercise independent judgment when artificial intelligence participates in creating, organizing and interpreting that knowledge.
That is an important evolution for ICARUS, and potentially for digital education more broadly.
Taking the question to UNESCO Digital Learning Week 2026
In September, ICARUS will participate in UNESCO Digital Learning Week 2026 at UNESCO Headquarters in Paris.
This year’s theme, “Education in the age of AI: Facts | Frictions | Frontiers,” could hardly be more relevant. UNESCO is placing human agency, equity and the public purpose of education at the centre of the discussion while examining synthetic knowledge, emerging AI systems and the new pressures they create for education.[4]
We intend to carry a simple question into that conversation:
Are we developing human capability as quickly as we are increasing machine capability?
The answer will matter well beyond education technology.
It will influence the quality of teaching, the integrity of assessment, the resilience of institutions and the ability of future citizens to navigate societies in which synthetic information becomes increasingly difficult to distinguish from reliable knowledge.
The next frontier is human
AI will continue to improve. Education systems should assume that future models will become faster, more autonomous and more convincing.
That makes human judgment more valuable, not less.
The next generation of AI education should therefore move beyond adoption metrics. The number of users tells us whether a technology has spread. It tells us very little about whether people can use it responsibly.
We should begin measuring something harder. Can teachers challenge an AI-generated answer? Can learners distinguish evidence from confident synthesis? Can institutions preserve human accountability when automation becomes easier? Can people decide when AI belongs in a learning process and when it does not?
At ICARUS, we believe these questions define the transition from AI access to AI agency.
Technology has already entered the classroom. Now education must make sure that human judgment enters with it.
Sources
- UNESCO — AI and Technologies in Education https://www.unesco.org/en/digital-education
- UNESCO — Survey: Two-Thirds of Higher Education Institutions Have or Are Developing Guidance on AI Use https://www.unesco.org/en/articles/unesco-survey-two-thirds-higher-education-institutions-have-or-are-developing-guidance-ai-use
- UNESCO — AI Competency Framework for Teachers https://www.unesco.org/en/articles/ai-competency-framework-teachers
- UNESCO Digital Library — AI Competency Framework for Teachers https://unesdoc.unesco.org/ark:/48223/pf0000391104
- UNESCO — Digital Learning Week 2026 https://www.unesco.org/en/weeks/digital-learning
- Official page for UNESCO Digital Learning Week 2026 and the theme “Education in the age of AI: Facts | Frictions | Frontiers.”UNESCO — Digital Learning Week 2026 & UNESCO ICT in Education Prize https://www.unesco.org/en/articles/digital-learning-week-2026-award-ceremony-2026-unesco-ict-education-prize
- UNESCO — Artificial Intelligence in Education https://www.unesco.org/en/digital-education/artificial-intelligence
- AI-Enabled Multilingual Learning Infrastructure for Inclusive Lifelong Learning https://www.unesco.org/sdg4education2030/en/knowledge-hub/ai-enabled-multilingual-learning-infrastructure-inclusive-lifelong-learning
- UNESCO — AI and the Futures of Learning https://www.unesco.org/en/digital-education/ai-future-learning students.