AI for Learning Outcomes: Equity, Safety, and System Transformation| AI Impact Summit 2026

Summary of AI Impact Summit 2026 Panel Discussion

---

Key Topics Covered

• AI adoption in education systems — moving from pilots to nationwide scaling
• Teacher empowerment and digital literacy — teachers as primary infrastructure for equity
• Global learning crisis — 70% of 10-year-olds in low/middle-income countries unable to read simple text
• Equity and digital divide — preventing AI from widening existing educational inequalities
• National sovereignty in AI development — small nations building independent language models and infrastructure
• Transparent, ethical AI use — learner consent and understanding of AI capabilities/risks
• System-level interoperability — breaking fragmentation of EdTech tools
• Research and evidence — measuring actual impact on learning outcomes vs. performance claims
• Teacher training and upskilling — cultural transformation within schools alongside technology rollout
• Resilience and offline capability — AI solutions that work in low-bandwidth environments

---

Key Points & Insights

1. Teachers are the infrastructure of equity — Not devices, connectivity, or free tools; rather, teacher empowerment and upskilling create measurable equity gains. One trained teacher can support 100+ students with AI-enabled instruction.

2. The three critical shifts required:

   • Move from pilots to system-level implementation
   • Shift from marketing claims to evidence-based learning outcomes
   • Build interoperable ecosystems that ministries can sustainably manage

3. Estonia's rapid rollout (within 1 year) succeeded through:

   • Universal simultaneous adoption (all schools at once, not regional pilots)
   • Deep public trust (built through previous digital success)
   • Pedagogical framework supporting technology (not technology first)
   • Whole-nation cooperation involving government, companies, and educators
   • Custom national language models (custom GPT for Estonian)

4. AI tools must be enablers, not replacements — AI should reduce teacher administrative burden, support differentiated instruction, and assess student progress—never replace teachers or students' learning process.

5. Global digital divide in AI governance is widening — Of 193 UN member states, only 7 participate meaningfully in major AI governance initiatives (GPAI, AI for Good, Hiroshima Process), leaving 119 nations excluded. This threatens to deepen the AI divide.

6. Low-bandwidth, offline-capable solutions are essential — AI solutions cannot assume high connectivity or compute availability; they must function in resource-constrained contexts to ensure true inclusion.

7. One-size-fits-all AI is insufficient — AI trained on California-centric values or English-only data marginalizes small languages and cultural contexts. Multilingual, locally developed models are necessary.

8. 2026 is a decision point, not a starting point — Students and teachers are already using AI independently. The choice is not whether AI enters schools, but whether institutions guide this ethically or cede control to unregulated use.

9. Transparency builds trust — Learners must know when AI is used, what it can/cannot do, and what risks it carries. This knowledge enables critical, responsible engagement rather than blind adoption.

10. University curricula need "best-before dates" — Educational content requires continuous renewal; a degree should not represent final competency. Reskilling every 2 years is becoming essential in the AI era.

---

Notable Quotes or Statements

President Alar Karis (Estonia):

"The question is not whether AI is used, but whether it is used knowingly, critically, and responsibly by everyone, not only by the most tech-savvy."

"In the AI era, it is not how smart machines are that matters most, but how smart the people who use them are. Education is the key to this."

Pia Lovrito (UNICEF):

"For three decades we've had EdTech. Devices did not create equity. Free tools did not create equity. Connectivity did not create equity. What created equity was an ecosystem that supports teachers, protects children, and is accountable to public systems."

"No generation before us and no generation after us bears this core responsibility. We have to bridge the digital divide before it gets wider."

"No more pilots. We have to work at a system level. We have to change and transform the system."

Mary Kerema (Kenya):

"We are even late. So we have to run... Students go to universities, lecturers are complaining they give assignments and students respond with ChatGPT. Do we sit and wait for our people to use technology in a wrong way? No. We need to up our game."

"How do I humanize technology? Use it to make me better, to make me more efficient, rather than fearing it."

Ivo Vissak (AI Leap, Estonia):

"A pedagogical program that's supported by technology, not a technological program supported by pedagogy."

"AI is both a good and bad technology for education. The bad reason is a very good reason: it's disruptive to a lot of very bad methods still entrenched in teaching practices."

Petri Mäki-Myllymäki (Finland):

"One size fits all does not work... If the value framework is Californian, that's only one value framework. In other parts of the world we have different values."

"Given an opportunity, people are willing to learn about this new technology... they should learn so that they can resist misinformation and all the hype."

---

Speakers & Organizations Mentioned

Government & Policy:

• President Alar Karis — President of the Republic of Estonia
• Mary Kerema — Kenyan Secretary for E-Government and Digital Economy
• Ivo Vissak — CEO, AI Leap (Estonia national AI education initiative)

International Organizations:

• UNICEF — represented by Pia Lovrito, Global Director of Education
• UNICEF Executive Board (Estonia as President this year)

Academic & Research:

• Professor Petri Mäki-Myllymäki — Finland, AI development and education
• Professor Jaan Aru — University of Tartu (Estonia), measuring AI impact on learning outcomes
• Professor Susanna Loeb — Stanford University, AI education impact research partner

Technology Companies (mentioned as partners):

• OpenAI — developed custom GPT for Estonian language education
• Google — core infrastructure provider in Estonian schools
• Microsoft — core infrastructure provider in Estonian schools

Other Initiatives:

• GPAI (Global Partnership on AI)
• AI for Good
• Hiroshima Process (AI governance)
• MasterCard Foundation — partnership with Kenya on AI for equalization

---

Technical Concepts & Resources

AI Models & Tools:

• Custom GPT (OpenAI) — Estonian language-specific model developed for learning outcomes delivery
• ChatGPT — widely used by students and cited as a catalyst for rethinking assessment methods
• Elements of AI — online course developed in Finland, translated to all EU countries, 2 million users globally; designed for citizens at large, not just university students

Infrastructure & Approaches:

• Large language models (LLMs) — capability to read and remember entire internet content; discussion of small language models as alternative to English-dominant systems
• National language models — emphasis on building small-language AI infrastructure to prevent linguistic marginalization
• Offline-capable AI systems — requirement for low-bandwidth and offline functionality to serve resource-constrained contexts
• Interoperable ecosystems — call for standardized, ministry-managed AI tool frameworks instead of fragmented app-based solutions

Measurement & Research:

• Learning outcome metrics — focus on knowledge retention, skill development (critical thinking, inquiry), and equity-specific measures
• Impact research frameworks — studying effects on:
  • Student learning outcomes
  • Teacher workload
  • Digital well-being
  • Equity gaps
  • Classroom practice quality
• Longitudinal studies — partnerships between University of Tartu and Stanford University to measure sustained impact

Policy Concepts:

• Data literacy and digital literacy as foundational skills — understanding decision-making, data use, rights in digital environments
• Teacher-centric AI design — AI as assistant to pedagogical practice, not replacement
• Transparent AI use protocols — learner consent and disclosure frameworks

---

Context & Background Notes

• Setting: AI Impact Summit 2026, Delhi, India (Bharat Mandapam venue)
• Timing: This discussion positions 2026 as a critical decision year—students and practitioners are already using AI; institutional resistance is the main barrier, not technical feasibility
• Key statistic: 70% of 10-year-olds in low- and middle-income countries lack basic reading comprehension; AI offers potential to scale teacher effectiveness if deployed equitably
• Governance gap: 119 of 193 UN member states are excluded from major global AI governance forums, limiting their voice in international AI policy