In the institutional rush to equip students and educators with AI competencies, we are committing a consequential pedagogical error: we are teaching skills before ethics. We are demonstrating how to prompt, how to generate, how to iterate - before we have established why responsible use matters, what the documented harms are, and where the normative boundaries should be drawn.

This sequencing is backwards. And in a K-12 context - where we are shaping the epistemic habits and moral reasoning of developing minds - it is genuinely dangerous.

I state this as someone who is unequivocally committed to AI integration in education. I founded iTeachAI Academy. I have facilitated AI professional development for educators across all fifty states, reaching over 1,250 course enrollments. I present on AI integration at FETC - in 2024, 2025, and 2026. I hold credentials as a Google for Education Certified Trainer and Coach, an Adobe Creative Educator, and an Apple Teacher. I am not anti-AI. But I am urgently, insistently pro-ethics. And in twenty-eight years of K-12 education, I have learned - repeatedly and sometimes painfully - that the sequence in which we teach matters as much as the content we teach.

The Empirical and Moral Case for Ethics First

Consider how we approach other powerful tools in educational settings. We do not hand a student a Bunsen burner and say, "Experiment with this - we will cover laboratory safety protocols next week." We do not give a teenager car keys and say, "Drive around the neighborhood - we will discuss traffic laws and defensive driving later." We establish the ethical and safety framework first, precisely because the tool is powerful enough to cause genuine harm.

Artificial intelligence is powerful enough to cause genuine harm. Not the speculative, science-fiction, sentient-robot variety. Real, measurable, empirically documented harm that is occurring in classrooms, communities, and digital ecosystems right now.

A 2025 systematic review published in ScienceDirect - analyzing sixty-eight peer-reviewed publications on K-12 AI ethics education from 2014 to 2025 - found that AI ethics education remains "significantly underprioritized in classroom practice" despite the global push for AI literacy curricula. The authors propose a competency-based responsible AI literacy framework that reconceptualizes AI ethics not as a supplementary unit but as a "transformative learning dimension" - a foundational orientation that must precede and permeate all AI skill instruction.

The documented harms demanding this ethics-first orientation are neither theoretical nor rare:

None of these harms require malicious intent. Every one of them can result from uninformed, well-meaning use by students and educators acting in good faith. That is precisely why ethics must come first - not as a box to check, but as a lens through which every subsequent AI interaction is filtered.

The Ethics-First Framework: Five Foundations Before Any Tool Training

In my iTeachAI Academy courses, I have developed what I call the Ethics-First Framework for AI education - a structured sequence of five ethical foundations that are established before any tool-specific training begins. This framework aligns with UNESCO's 2024 AI Competency Framework for Teachers, which positions "Ethics of AI" as one of its five foundational competency dimensions, and with the emerging state-level guidance documents that now exist in over forty states.

Foundation 1: Radical Transparency

The principle: Always disclose when and how AI has contributed to your work.

Why it matters: Transparency is the epistemic bedrock of academic integrity and intellectual trust. In an educational context, students must understand that using AI without disclosure is not merely a rule violation - it is a form of epistemic dishonesty that misrepresents their competencies, undermines the developmental purpose of learning tasks, and erodes the trust that makes formative assessment meaningful.

How I teach it: I have students use AI to generate a piece of analytical writing, then compose a detailed "process statement" documenting exactly how AI was used, what they modified, what they rejected, and what original intellectual labor they contributed. This practice simultaneously normalizes transparency, develops metacognitive awareness, and teaches students to maintain authorial agency even when working with AI tools. Several states - including Missouri's 2025-2026 AI guidance for local education agencies - now recommend similar transparency protocols as part of district-level AI policy.

Foundation 2: Epistemic Vigilance

The principle: Never accept AI outputs without independent verification and critical evaluation.

Why it matters: Generative AI systems produce text, data visualizations, and images that possess the surface markers of authority - confident tone, structured argumentation, citation-like formatting - while potentially containing factual inaccuracies, logical fallacies, statistical fabrications, and encoded biases. The appearance of epistemic authority is precisely what makes uncritical acceptance so dangerous, particularly for students who are still developing their capacity for source evaluation and evidential reasoning.

How I teach it: I provide students with AI-generated content that contains deliberately embedded errors - factual inaccuracies, circular reasoning, biased framing, fabricated citations - and challenge them to identify the problems. This exercise builds the habit of verification, demonstrates concretely that AI outputs require human judgment, and develops transferable critical thinking skills that extend far beyond AI-specific contexts. Students invariably discover that their initial confidence in their ability to detect AI errors was misplaced - a humbling but pedagogically powerful realization.

Foundation 3: Data Sovereignty and Privacy Awareness

The principle: Understand what happens to the data you share with AI systems - and protect yourself and others accordingly.

Why it matters: Every interaction with an AI tool constitutes a data transaction. Students and educators must understand what data they are sharing, how it may be stored, processed, and monetized, and what the privacy implications are - particularly for minors whose data enjoys heightened legal protections under FERPA, COPPA, and emerging state privacy statutes. California's AI literacy guidance explicitly emphasizes that AI education must be embedded across content areas and must include data privacy awareness from the earliest elementary grades.

How I teach it: I walk through the terms of service and privacy policies of widely used AI tools - translated into age-appropriate language - and guide students in identifying what data each platform collects, retains, and potentially shares. This exercise is consistently eye-opening: students are often shocked to learn the scope of data collection embedded in tools they use casually. I establish non-negotiable classroom protocols: never share your real name, home address, phone number, school name, or other personally identifying information with any AI system. These protocols are not restrictions on learning - they are preconditions for safe learning.

Foundation 4: Systematic Bias Recognition

The principle: AI systems reflect and amplify the biases present in their training data. Learn to recognize, name, and challenge those biases.

Why it matters: AI tools deployed in educational contexts can perpetuate stereotypes, underrepresent marginalized communities, reinforce systemic inequities, and produce outputs that normalize discrimination - in hiring simulations, content generation, image creation, language processing, and recommendation algorithms. Students who do not possess the analytical tools to recognize these biases are vulnerable to absorbing them as objective truth.

How I teach it: I employ comparative analysis exercises in which students generate AI content about different demographic groups, professional roles, or cultural contexts - then systematically analyze the outputs for patterns of bias. Asking an AI image generator to create pictures of "a doctor," "a nurse," "a CEO," "a teacher," and "a criminal" - then examining the gender, racial, age, and socioeconomic patterns in the outputs - generates data that speaks for itself and catalyzes some of the most substantive classroom discussions I have ever facilitated. The SREB's 2025 guidance framework recommends precisely this kind of structured bias analysis as a core component of K-12 AI literacy education.

Foundation 5: Intentional Human Agency

The principle: You are always the decision-maker. AI is the instrument; you are the author, the evaluator, and the moral agent.

Why it matters: There is a subtle but empirically documented risk that students - and adults - begin to defer to AI: accepting its recommendations without evaluation, following its suggestions without reflection, allowing it to shape their thinking rather than deploying it in service of their own intellectual goals. UNESCO's AI Competency Framework for Teachers identifies "a human-centred mindset" - encompassing human agency, accountability, and social responsibility - as its foundational competency dimension. Maintaining human agency means using AI as an instrument in service of human purposes, never permitting AI to define those purposes.

How I teach it: I design every AI-integrated activity so that the student must exercise final judgment. AI generates three thesis statements; the student selects one and writes a justification defending the choice. AI proposes a research methodology; the student evaluates it against their existing knowledge and decides whether to adopt, modify, or reject it. AI drafts an argument; the student identifies its strongest and weakest elements and reconstructs it in their own voice. In every case, the AI occupies the role of assistant - never authority. This structural design choice is pedagogically intentional: it habituates students to a relationship with AI that preserves and strengthens their intellectual autonomy.

Why Schools Resist Ethics-First - And Why They Must Not

I will be candid: the ethics-first approach encounters resistance, and not always from the constituencies one might expect.

Administrators sometimes resist because they are under institutional pressure to demonstrate visible AI adoption rapidly. Ethics conversations feel slow when board members and community stakeholders are asking what the district is doing about AI. My response is empirical and pragmatic: moving fast without an ethical foundation produces crises that halt all forward progress - academic integrity scandals, data privacy incidents, community backlash. The ethics-first approach is actually faster in aggregate because it prevents the policy emergencies that invariably arise from uninformed, unscaffolded adoption. Ohio's House Bill 96, which mandates formal AI policies in every district by July 2026, reflects the growing recognition that governance must precede adoption.

Some educators resist because they want to reach the practical applications - the tools, the workflow efficiencies, the time-saving techniques. I understand this urgency viscerally. Teachers are stretched impossibly thin, and they want solutions, not what they perceive as philosophical preamble. My approach is to make the ethics inherently practical. The bias-recognition exercise is simultaneously a critical thinking lesson aligned to ELA and social studies standards. The transparency practice is simultaneously a metacognitive writing process lesson. The privacy audit is simultaneously a digital citizenship lesson. Ethics-first does not mean skills-never - it means skills are taught within an ethical context from the first moment, and every skill lesson reinforces the ethical orientation.

Even some students resist, particularly older adolescents who have been using AI tools independently for months and perceive the ethics instruction as remedial or paternalistic. For these students, I deploy what I call the epistemic humility exercise: I present AI-generated content that is convincingly, artfully, comprehensively wrong - and ask them to evaluate its accuracy. The moment they discover they were fooled - that their confidence in their own AI discernment was unwarranted - the ethics conversation transforms from abstract to urgent, from imposed to personal. It is consistently one of the most powerful pedagogical moments in my courses.

Integration, Not Segregation

The objective of the Ethics-First Framework is emphatically not to create a standalone ethics unit that sits cordoned off from AI skill instruction - a quarantined module completed and forgotten before the "real" learning begins. The objective is to establish an ethical orientation so deeply integrated that it permeates every AI interaction, every assignment, every evaluation.

In my iTeachAI courses, every tool demonstration incorporates an ethics checkpoint. Every AI-integrated assignment includes a transparency requirement. Every evaluation of AI output includes a bias analysis component. The ethics are not ancillary to the skills - they are woven into the fabric of every skill, inseparable and mutually reinforcing.

This integrated approach reflects a lesson the education community should have learned - but largely did not - from the digital literacy movement of the early 2000s. We taught students to use the internet, then spent the subsequent decade managing the consequences of cyberbullying, misinformation proliferation, and digital addiction. We built the skills without the ethical scaffolding - and we are still paying the price.

With artificial intelligence, we have a rare second chance. A chance to get the sequencing right. A chance to build the ethical foundation before the structural weight of widespread adoption makes retrofitting impossible.

Ethics first. Skills always. In that order. Without exception.

Janette Camacho, Ed.D. is the founder of iTeachAI Academy, a Google for Education Certified Trainer and Coach, a FETC 2024/2025/2026 Featured Presenter, an Adobe Creative Educator, an Apple Teacher, and an EdTech Digest 2026 Honoree. She advocates for ethics-centered AI education nationwide, reaching educators across all 50 states through iTeachAI Academy's 1,250+ course enrollments.