AI Transparency

Last updated: May 2026. Updated when the AI provider, models, or data handling terms change.

GradeMap uses artificial intelligence to support your study. This page explains how, what data is involved, and what the limitations are. We publish this so you, your institution, and your educators can make informed decisions about using the platform.

How GradeMap uses AI

GradeMap uses AI as a study coach, not a ghostwriter. Every interaction is designed to help you understand and engage with your subject — not to produce content you submit as your own.

The AI is used for:

Socratic coaching — asking questions to help you understand concepts rather than explaining them outright
Rubric analysis — translating academic rubric criteria into plain language and identifying what distinguishes grade bands
Writing feedback — reviewing your draft text for academic tone, clarity, and structure without rewriting it for you
Citation and referencing — generating correctly formatted references and auditing your reference list
Assignment planning — helping you break down an assessment into a concrete starting structure and first steps
Research assistance — analysing sources you provide for relevance to your assessment criteria

GradeMap does not write assignments for you. All coaching sessions are structured to keep you as the author of your own work. Feature toggles let you and your institution control the level of AI assistance.

AI provider and models

GradeMap is powered by Anthropic’s Claude family of models. Different models are used for different task types based on the complexity and cost of each interaction.

Model	Used for
`claude-haiku-4-5-20251001`	Simple confirmations and session logging
`claude-sonnet-4-6`	Coaching, rubric analysis, and writing feedback
`claude-opus-4-7`	Citation auditing and referencing

Students can optionally provide their own Anthropic API key (BYO Key), in which case their coaching sessions run against their personal account. The coaching methodology and system prompts remain server-side in all cases.

Training data and knowledge limits

Claude’s training data has a cutoff of approximately early 2025. This means:

Recent publications, legislation, court decisions, and research findings may be incomplete or missing
Referencing formats reflect the standard as of the training cutoff; major style guide revisions after that date may not be reflected
University policies, course outlines, and assessment criteria are always taken from content you paste in — the AI does not retrieve them independently

Always verify time-sensitive facts, recent statistics, and citation details from primary sources.

Your content and AI training

GradeMap does not use your subject content or coaching conversations to train AI models.

Content you enter — subject outlines, rubrics, draft text, research sources — is sent to Anthropic’s API solely to generate your responses. Under our data processing agreement with Anthropic, this content is not retained by Anthropic for training purposes.

If you use a BYO API key, data processing is governed by your own Anthropic account terms. Review your account settings at console.anthropic.com to confirm your data handling preferences.

Refer to Anthropic’s Privacy Policy and Anthropic’s Usage Policy for the authoritative terms.

Research integrity

GradeMap is built around the principle that AI assistance should support your learning, not replace it. The coaching methodology enforces this in several ways:

Coaching uses Socratic questioning — the AI asks you questions rather than telling you answers
Writing feedback identifies issues for you to fix; the AI does not rewrite your text
Assignment kickstarters produce a structure and first steps, not a draft
Referencing assistance helps you cite correctly; it does not fabricate sources
If you paste your university’s AI use policy into your profile, GradeMap will adapt its coaching to stay within those guidelines

GradeMap is designed for use within Australian university academic integrity frameworks. Students are responsible for understanding their institution’s specific AI use policy and for complying with it when submitting work.

For a principle-by-principle review of GradeMap against the Australian Code for the Responsible Conduct of Research 2018, see the ACRR 2018 Alignment Statement.

Limitations and bias

AI outputs may reflect biases present in training data. Known limitations include:

Cultural and linguistic bias — Claude was primarily trained on English-language content; responses may not fully reflect non-Western academic traditions or perspectives
Discipline conventions — coaching is grounded in your pasted subject content, reducing generic responses, but discipline-specific conventions may not always be accurate
Confident errors — AI can produce plausible-sounding but incorrect information, particularly for citations, statistics, and factual claims. Verify all factual content independently
Referencing accuracy — the referencing agent is designed for near-infallible output, but always verify against the original source before submitting

GradeMap mitigates these risks by grounding coaching responses in your own subject content and by routing higher-stakes tasks (citation auditing) to the most capable model available.

University AI policies

Many Australian universities have published AI use policies for students. GradeMap supports compliance with these policies in two ways:

Policy adaptation — paste your university’s AI policy into your profile and GradeMap will extract the key constraints and apply them to all coaching interactions
Feature toggles — AI assistance levels can be configured per user, allowing students to operate within the bounds their institution or integrity office specifies

If you represent a university or integrity team and have questions about how GradeMap operates, contact us at [email protected].

Your data rights

GradeMap is designed with data rights built in from the start:

Export — you can download a complete export of all your data at any time from Settings
Deletion — you can request full account deletion from Settings; all personal data is removed within 30 days
Portability — session exports are in Markdown format, making them readable outside GradeMap

For the full privacy and data handling terms, read the GradeMap Privacy Policy.