How Luna uses AI
Luna uses artificial intelligence to help disabled and neurodivergent students study more effectively. Luna's AI generates flashcards exclusively from content provided by the student – whether that is notes, lecture slides, reading material, transcripts, or other learning resources. It does not write academic content, introduce external knowledge, or complete any work on the student's behalf. This page explains how Luna's AI works, what it does and does not do, and how student data is protected.
SECTION 1
Our core principle
Luna's AI is a compression and retrieval tool, not a content generation tool.
When a student provides learning material, Luna's AI reorganises that existing information into flashcards. The student remains the author of every idea being studied. It does not write, create, or invent academic content.
Three rules govern every flashcard Luna generates:
Student Authorship Rule. Every flashcard must be derived from content provided by the student. Luna does not generate new subject knowledge or introduce information from outside the source material.
No Supplementing Rule. Luna does not add context, explanation, or background not present in the student's source material.
No Correction Rule. Luna does not fix inaccuracies or silently change meaning within the student’s source material.
SECTION 2
What Luna’s AI does
Structures students' learning materials: Luna's AI identifies key concepts from content provided by the student and formats them as retrievable flashcards. This removes the effort of manually converting learning materials into study resources.
Builds in metacognitive checkpoints: Luna's AI generates flashcards directly from the student's learning materials, not from external sources or general knowledge. This means students can focus on learning rather than verifying, correcting, or second-guessing AI-generated content. Built-in review prompts give students a natural moment to sense-check before moving forward – keeping them in control of their cards without breaking their concentration.
Generates one fact per card: Luna's AI creates single-concept flashcards rather than compound or multi-part cards. For neurodivergent students, particularly those with ADHD, working memory differences, or processing difficulties, reducing each card to a single retrievable fact lowers cognitive load and improves retention. This is also an academic integrity feature: a card containing one discrete fact drawn from the student's learning materials cannot be mistaken for AI-generated content or synthesis.
Schedules review intelligently: Luna puts students in control of the study schedule. It gives the option to use spaced repetition to surface the right card at the right time, based on each student's own rating of their confidence.
SECTION 3
What Luna's AI does not do
Write academic content: Luna does not generate essays, write answers, or produce any content that could be submitted as the student's own assessable work.
Introduce external knowledge: Flashcard content is drawn exclusively from material provided by the student. This may include notes, lecture slides, transcripts, reading material, or other learning resources. Nothing is sourced from the web or a general knowledge model. If a student submits only a question or topic heading with no factual content, Luna produces no output. The only time that Luna introduces concepts from outside the student's material is to create plausible distractors for true-or-false questions – and even then only if there is nothing that can be used within the source.
Correct students' learning materials: Luna does not edit, verify, or fact-check what a student has written. Accuracy of the source material remains the student's responsibility throughout.
SECTION 4
Academic integrity
University AI policies distinguish between tools that assist learning and tools that complete assessed work on the student's behalf. Luna falls clearly into the first category.
Because Luna's AI only processes content provided by the student, it cannot generate new subject knowledge or create original academic work. The student remains responsible for selecting, understanding, and studying the material. Luna helps organise that material for revision – it does not replace learning.
SECTION 5
Student data and privacy
When a student submits learning materials for flashcard generation, that text is processed by Luna's AI provider via a secure API. Only the content the student enters into the learning material field is sent – no personally identifiable information such as name, email address, or account details is transmitted. Luna does not send any data held elsewhere in the product.
Luna only works with AI providers whose API terms explicitly exclude customer data from model training by default. This is a condition of selection, not an assumption. Luna has not opted in to any data sharing arrangement with its current or previous AI providers.
Luna does not retain students' original source material after flashcard generation. The key facts and insights extracted during that process are stored within the student's collection and remain there until the student chooses to delete it. No original source text is held by Luna beyond the point of processing.
SECTION 6
Accessibility of AI features
Luna's AI features are built within the same accessible interface that neurodivergent and disabled students use every day – not a separate layer bolted on.
For many neurodivergent students, having to switch between multiple tools is itself an accessibility barrier. Having AI assistance within a purpose-built, low-friction, accessible environment is qualitatively different from being directed to a general-purpose AI tool.
SECTION 7
Common questions
Does Luna's AI breach academic integrity policies?
No. Luna generates flashcards only from content the student has already produced. It is trained not to generate subject knowledge, write assessable work, or introduce content the student has not authored. This is structurally equivalent to a highlighter or a screen reader – it helps the student access and review their own knowledge more effectively.
Will Luna correct my learning materials?
No. Luna does not edit, verify, or fact-check what a student has written. The accuracy of source material remains the student's responsibility. Luna does prompt students to review their learning materials before upload and to check generated flashcards before they are built – keeping the student in control throughout.
Does Luna just do the work for the student?
Luna helps students turn existing learning material into revision resources more quickly and efficiently. The student still needs to engage with lectures, reading, course materials, and independent study before there is anything for Luna to work from. Luna supports revision – it does not replace the learning process.
Why does Luna create one fact per flashcard?
Research consistently shows that reducing the information load per retrieval prompt improves both retention and recall accuracy for students with working memory differences, ADHD, and processing difficulties. A single-fact card isolates exactly what the student needs to remember and reduces the cognitive load of interpreting a multi-part question.
Will student learning materials be used to train AI models?
No. Luna only works with AI providers whose API terms exclude customer data from model training by default. Luna has not opted in to any data sharing arrangement. As with all API platforms, providers may retain limited logs for abuse monitoring and platform security purposes – this is standard practice and is distinct from model training. Luna's selection criteria require that this default position is maintained.
How does Luna decide which AI model to use?
Luna reviews available AI models on an ongoing basis and selects the provider best placed to enforce its core rules – student authorship, no supplementing, no correction – at any given time. Our obligation is to those rules, not to any particular vendor. As the AI landscape evolves, the model may change. The constraints it operates under do not. Responsible AI selection means choosing the best available tool for the job and reviewing that decision continuously – not locking to a single provider regardless of how the technology develops.