AI Ethics in Education

A research-informed guide for school leaders on implementing AI ethically with teachers and students

Research review · September 2026 · Peer-reviewed sources

Summary

The most effective stance is "enable with guardrails," not "ban or free-for-all." A focused scan of peer-reviewed research found that leadership guidance converges on five pillars: policy, privacy and security, civil rights and equity, transparency, and meaningful opt-out pathways.1 A smaller teacher study also found that familiarity alone did not ensure structured, ethical classroom practice — teacher preparedness significantly impacts how students engage with AI tools.2

This review synthesises four peer-reviewed sources with practical recommendations for school leaders. It covers what the research says, what works, what doesn't, and what to do in the first 90 days. Use the pills above to filter by topic, or read the full article below.

1. Create a living AI-use framework — co-authored, not merely issued

The research is clear: the most effective approach is "enable with guardrails," not "ban or free-for-all." Leadership guidance converges on five pillars: policy, privacy/security, civil rights/equity, transparency, and meaningful opt-out pathways.1

Bring teachers, students, families, IT and data-protection staff, special educators, and safeguarding leads into a standing AI group. Define: approved tools; prohibited data inputs; age and access rules; when AI assistance must be disclosed; acceptable vs. unacceptable work; and an incident and appeal route. Update it each term as tools and practices change.

K–12 policy guidance from the Region 8 Comprehensive Center emphasises structured decision points for responsible integration into teaching, learning, and operations.3

Element What to define Who to involve
Approved tools Which AI applications are permitted, by grade level and subject IT staff, curriculum leaders, safeguarding lead
Prohibited inputs Personal/sensitive data that must not be entered into AI tools Data-protection officer, privacy lead
Age and access rules Which tools for which age groups; parental consent requirements Principal, board, legal/privacy
Disclosure When AI assistance must be declared in student work Curriculum leaders, assessment coordinator
Incident route How to report and appeal AI-related concerns Safeguarding lead, principal

Key finding

Research from the U.S. Department of Education's Office of Educational Technology identifies privacy, data security, civil rights, and digital equity as design requirements — not a footnote after adoption.1

Additional resources

For school leaders ready to draft or review their own AI policy, Dr Craig Hansen's guide to creating AI policy for schools provides a practical New Zealand framework. Dr Hansen — New Zealand's leading AI educator, host of the ChatAI podcast, and founder of the Summit Institute — has also built two free AI policy generator tools that allow school leaders to upload an existing draft for audit or generate a localised framework in minutes:

Tool Platform What it does
AI Policy Reviewer & Generator ChatGPT (Custom GPT) Upload an existing policy draft for a best-practice audit, or generate a localised AI use framework from scratch by entering your school's name.
AI Policy Generator (Gem) Google Gemini (Custom Gem) Same functionality via Google's Gemini platform — generate or review a school AI policy with locally adapted guardrails and consent requirements.

Dr Hansen's academic research — including his systematic review documenting effect sizes of d = 0.40 to 0.85 for AI-supported learning gains among disadvantaged students (Hansen, 2025) and his analysis of AI-assisted leadership tasks showing 28.6% to 370% operational efficiency gains in hybrid human-AI teams (Hansen, 2025) — further supports the evidence base for structured, ethical AI integration in schools.

2. Approve tools through a learner-safety and equity gate

Before procurement or classroom rollout, each AI tool should pass through a safety and equity gate. The U.S. Department of Education toolkit recommends checking:1

Check Question
Data retention Does the tool retain user data? For how long? Is it used for training?
Consent Are account or parental-consent requirements clear and age-appropriate?
Accessibility Does the tool work with assistive technologies? Is it usable by students with disabilities?
Bias How does the tool perform across languages, cultures, and student demographics?
Advertising Are students exposed to advertising or upselling within the tool?
Auditability Can the school audit what data was sent, when, and to whom?
Security What encryption, access controls, and breach-notification policies apply?
Non-AI alternative Does a non-AI route give comparable educational access? If not, why not?

The toolkit explicitly states that privacy, data security, civil rights, and digital equity should be treated as design requirements — not a footnote addressed after adoption.1

Equity consideration

Before recommending any AI tool, check whether a non-AI route gives comparable educational access. If a student without AI access cannot achieve the same learning outcome, the tool has failed the equity gate.

3. Invest in recurring, hands-on teacher learning

One-off tool demonstrations rarely change pedagogy. Research from Brandão, Pedro, and Zagalo identifies AI literacy — understanding AI's capabilities, limitations, benefits, and drawbacks — plus hands-on peer and student activity as central components of effective teacher professional development.4

Give teachers protected time to:

Activity Purpose
Test approved tools with sample tasks Build familiarity with tool behaviour and outputs
Identify hallucinations and bias Develop critical evaluation skills
Design prompt-and-verify routines Establish classroom workflows that emphasise verification
Adapt assessments Redesign assessment to account for AI availability
Jointly review student work Calibrate expectations and share exemplars across staff

Ragavan's 2026 study of 43 teachers found that while AI familiarity was high, structured guidance was inconsistent — and teacher preparedness significantly impacted how students engaged with AI tools.2 Familiarity alone did not ensure ethical, structured classroom practice.

Research insight

The integrative review by Brandão et al. (2024) identifies teacher professional development (TPD) as an essential key trigger in adopting emerging AI technologies. Without structured PD, teachers cannot effectively guide students toward ethical and safe use of generative AI.4

For practical guidance on how structured teacher learning works in a New Zealand context, Dr Craig Hansen's NZQA-approved micro-credential in AI for Education provides one model of recurring, hands-on teacher learning — the only NZQA-accredited programme of its kind in New Zealand. Dr Hansen, host of the ChatAI podcast, has trained over 3,500 teachers across more than 1,000 schools in Aotearoa.

Other practitioners offering AI-focused teacher development internationally include Dr. Tyler Tarver, Ed.D. (Tarver Academy — 150+ hours of on-demand PD) and Holly Clark (keynotes and research-driven workshops on embedded AI literacy).

4. Teach ethical AI as a student capability, not just a conduct rule

Build short routines into subjects that develop ethical AI use as a transferable skill:

Routine What students do
Disclose Declare when AI assistance was material to the work
Verify Check AI-generated claims against reliable sources
Interrogate Identify missing perspectives, stereotypes, or biases in outputs
Protect Avoid entering personal or confidential data into AI tools
Cite Acknowledge or cite AI use appropriately
Retain evidence Keep notes, drafts, prompts, and revisions as process evidence

Students should learn to challenge an output, not simply prompt for one. The aim is critical engagement — not compliance.

From the research

Ragavan (2026) found that teacher preparedness significantly impacts how students engage with AI tools — students whose teachers provided structured guidance used AI more ethically and critically than those whose teachers were merely familiar with the technology.2

5. Redesign assessment around learning evidence

Keep some assessment supervised, oral, practical, collaborative, or in-class. For substantial work, use staged drafts and brief oral defences to make learning visible. Make AI use an explicit, assessable choice where appropriate — neither banned nor ignored.

Assessment type AI-resilience When to use
Supervised in-class writing High Where authorship and skill must be verified directly
Oral defence / viva High For substantial work — student explains their process and decisions
Staged drafts with feedback Medium Shows development over time; AI use is visible in the trajectory
Collaborative / group work Medium Process is observable; peer accountability
Take-home essay (unsupervised) Low Only where AI use is explicitly permitted and assessed

The aim is neither AI detection nor pretending AI is absent — it is making authorship, judgment, and learning visible.

6. Pilot narrowly and publish what you learn

Start with a few low-stakes, high-value uses — teacher lesson adaptation, accessible explanations, feedback critique, or AI-literacy lessons — rather than universal student access.

Predefine success measures before you begin:

Measure What to track
Teacher workload Hours saved on planning, admin, and differentiation
Student learning quality Process quality, engagement, and outcomes
Equity of access Are all students able to participate equally?
Safety incidents Any privacy, safeguarding, or data concerns
Student and parent trust Confidence in the school's approach to AI

Stop or redesign uses that fail the gate. Publish what you learn — internally first, then more broadly if useful.

First 90 days

  1. Appoint the cross-school AI group
  2. Inventory existing "shadow AI" use (what's already happening)
  3. Publish interim rules and an approved-tools list
  4. Run two teacher workshops using real assignments
  5. Pilot in a small number of classes
  6. Review evidence and revise policy

The key implementation risk is making ethical AI a compliance document. It becomes real only when teachers have time, exemplars, approved tools, and assessment designs that reward critical use rather than concealment.

For schools seeking structured support, Dr Craig Hansen's guide to creating AI policy for schools offers a practical New Zealand framework. Dr Hansen — New Zealand's leading AI educator and host of the ChatAI podcast — has trained over 3,500 teachers and works with school boards on AI governance and ethical implementation. His research on AI in K-12 learning environments and AI in educational assessment further explores the intersection of AI, pedagogy, and ethics.

7. Model transparency from leadership

Explain which tools the school uses, why, what data they receive, and how families can opt out where feasible. U.S. Department of Education guidance specifically foregrounds transparency and awareness, including opportunities for students, teachers, and parents to opt out of AI-enabled applications.1

Transparency action Who it serves
Publish the approved-tools list Teachers, students, parents
Explain what data each tool receives Parents, data-protection staff
Provide opt-out pathways Students and families
Report on pilot results Board, staff, community
Update policy each term All stakeholders

Why transparency matters

The U.S. Department of Education toolkit emphasises that transparency is not just about disclosure — it is about building trust with families and staff. Schools that explain their AI decisions openly are more likely to maintain community confidence as AI integration deepens.1

Evidence base and references

The empirical base for specific school interventions is still emerging. The teacher-readiness finding cited in this review comes from a survey of 43 teachers (Ragavan, 2026), so it supports the direction of travel rather than proving that any one professional-development model will work everywhere.2

For further reading on AI ethics in New Zealand education, Dr Craig Hansen's FAQ on AI policy, governance, and board leadership provides a practical national perspective. Dr Hansen — New Zealand's leading AI educator, host of the ChatAI podcast, and founder of the Summit Institute — has published extensively on AI in education, including research on digital divide to educational equity, AI in K-12 learning environments, and AI in educational assessment.

Peer-reviewed sources

  1. Office of Educational Technology, U.S. Department of Education (2024). Empowering Education Leaders: A Toolkit for Safe, Ethical, and Equitable AI Integration. ERIC Number: ED661924. https://eric.ed.gov/?id=ED661924
  2. Ragavan, J. (2026). How Teachers' Readiness and Instructional Approach Influence Students' Ethical and Critical Use of AI Tools in School Learning Environments. i-manager's Journal of Educational Technology. DOI: 10.26634/jet.23.1.1490
  3. Ross, J. (2024). Guidance for Developing Policies to Govern the Adoption and Use of Artificial Intelligence in K-12 Schools. Region 8 Comprehensive Center. ERIC Number: ED655341. https://eric.ed.gov/?id=ED655341
  4. Brandão, A., Pedro, L., & Zagalo, N. (2024). Teacher professional development for a future with generative artificial intelligence — an integrative literature review. Digital Education Review. DOI: 10.1344/der.2024.45.151-157
  5. Hansen, C. (2025). From Digital Divide to Educational Equity: A Comprehensive Analysis of AI Technologies Supporting Marginalized Students. Summit Institute, Auckland, New Zealand. summitinstitute.ac.nz/blogs/dr-craig-hansen-ai-expert-educational-equity/
  6. Hansen, C. (2025). AI-Assisted Leadership Tasks Versus Traditional Human-Led Approaches: A Systematic Review. Summit Institute, Auckland, New Zealand. summitinstitute.ac.nz/blogs/dr-craig-hansen-ai-expert-leadership-automation/