AI & Educational Use Policy

Introduction

Xmentra is a language-learning platform. This Policy explains how AI-assisted technologies are used in that work today, what they may be used for as the Services grow, where their limits sit, and what stays with you as a learner.

It is written to be read, not endured. It describes practice rather than promises, and it separates what happens today from what may happen later.

This Policy owns the description of AI capability and its limits. It forms part of the Terms of Use (/terms), which own the contractual terms, liability, and governing law. Processing of Personal Information is owned by the Privacy Policy (/privacy). Educational scope, outcome limits, and professional-advice boundaries are owned by the Disclaimer (/disclaimer). Rules about permitted and prohibited conduct, including misuse of AI Features, are owned by the Acceptable Use Policy (/acceptable-use).

1. Definitions

These definitions are shared across all Xmentra legal documents and are reproduced without modification.

• Platform — the Xmentra website at https://xmentra.com, together with any applications, subdomains, and interfaces operated by Xmentra.
• Services — all functionality offered through the Platform, including lessons, learning experiences, vocabulary, grammar, listening, revision, progress tracking, the educational blog, and any features added later.
• AI Features — any part of the Services that uses automated language models or similar techniques to generate, adapt, evaluate, or respond to content, including the AI Mentor and AI Conversations when introduced.
• Content — any text, audio, image, code, exercise, translation, or other material available through the Platform, whether created by Xmentra, generated by AI Features, or submitted by a User.
• Educational Content — Content published by Xmentra for learning purposes, including lessons, learning experiences, explanations, audio, and blog articles.
• Mentor — the guiding instructional voice of the Platform, delivered through Educational Content and, where applicable, AI Features. A Mentor is not a certified teacher, examiner, translator, or adviser.
• Learning Data — information about a User's learning activity, such as lessons opened, experiences completed, confidence responses, saved vocabulary, and revision history.
• Personal Information — information that identifies, or can reasonably be used to identify, an individual, either alone or combined with other information available to Xmentra.
• Local Storage — browser storage mechanisms (including localStorage and sessionStorage) used by the Platform to keep information on the User's own device.
• Third Party Services — independent providers whose infrastructure or tools support the Platform, including hosting, analytics, and, where applicable, payment, communication, and AI providers.
• User — any person who accesses or uses the Platform, with or without an Account.
• Account — a credentialed profile that a User may create when Accounts are available, allowing Learning Data to be stored by Xmentra rather than only on the device.
• Premium Features — Services available only under a Subscription or other paid plan.
• Community — any feature allowing Users to interact with one another, including forums, comments, groups, or shared practice.

2. Purpose of this Policy

Xmentra uses, and may further use, AI-assisted technologies for educational reasons only. The purposes are:

• supporting language learning with clear, usable material;
• helping learners practise language they will actually meet;
• generating or adapting learning material during content production;
• providing educational feedback, where such feedback is offered;
• assisting with learning pathways and what to revisit;
• supporting personalised learning experiences, where offered;
• improving the usability and clarity of the Services.

Some of these purposes describe how the Services work today; others describe capability that may be introduced. Section 4 states the difference plainly, and every later section is worded so that a conditional phrase — "where offered", "where AI Features are available" — marks anything not yet live.

3. Core AI Governance Principles

Three permanent principles apply to every AI Feature. They are shared across the Xmentra legal documents and are not restated differently anywhere else.

Transparency. Users are told when AI is involved. In practice this means the Platform describes AI-assisted functionality in this Policy, and where an AI Feature responds to you directly, the Services indicate that you are interacting with an AI Feature rather than a person.

Human oversight. AI assists judgement; it does not replace it. In practice this means Educational Content produced with AI assistance goes through human review before publication, that AI Features are designed and configured by people, and that no learner is graded, gated, or certified by a model. Section 9 explains the scope of that oversight honestly, including what it does not cover.

Educational assistance. AI supports learning and may be inaccurate. In practice this means AI-assisted output is offered as teaching help, is open to question, and should be checked wherever a decision depends on it.

4. What Is AI-Assisted Today

Current position, as of the Last Updated date above:

• Educational Content on the Platform is authored material. The Mentor voice you read in lessons and learning experiences is written and reviewed by people, with AI assistance used during production. It is not model output generated at the moment you open a lesson.
• No AI Feature currently processes text or audio you submit in order to generate a live reply to you.
• Learning Data is kept in Local Storage on your own device, as described in the Privacy Policy (/privacy) and the Cookie Policy (/cookie-policy). It is not sent to an AI provider by the current Services.

Capability that may be introduced later, and which this Policy already governs:

• an AI Mentor or AI Conversations that respond to what you write or say;
• AI-assisted feedback on your practice;
• AI-assisted personalisation of what you are shown next;
• AI-assisted generation or adaptation of exercises at the time of use.

Where any of these becomes available, this Policy and the Privacy Policy are updated in the same release, before or at the time the feature becomes active.

5. AI Is Assistance, Not Authority

AI Features provide educational assistance. They are a way of helping you learn, and nothing more than that.

AI-assisted output is not, and is not presented as:

• infallible or guaranteed correct;
• an authoritative statement about a language;
• a substitute for your own judgement;
• an official examiner, or an indication of how an examiner would score you;
• an official or certified translator;
• an accredited teacher or accredited educational provider;
• a regulated professional adviser of any kind.

This is consistent with the Disclaimer (/disclaimer), which owns the broader description of what the Platform does and does not promise.

6. Limitations of AI-Assisted Output

AI-assisted or AI-generated output may:

• misunderstand the context of a question;
• contain inaccuracies;
• omit information that matters to your situation;
• reflect information that is out of date;
• oversimplify something that is genuinely more complex;
• produce language that is grammatical but unnatural;
• produce wording that is technically correct yet inappropriate for the situation;
• read fluently and confidently while being wrong;
• vary between requests, so that the same question produces different answers.

None of this makes AI-assisted learning unsafe. It means the sensible posture is informed use: take the help, keep your judgement, and verify anything that carries weight.

7. Language-Specific Limitations

Language is not a single fixed system, and this matters more here than in most places where AI is used. AI-assisted language guidance may not account fully for:

• regional differences across countries and areas where a language is spoken;
• dialects and local usage;
• register, including how formal or informal a phrase sounds;
• the distinction between formal and informal address, and when each is expected;
• cultural context, and what is polite or unusual in a given setting;
• idioms, humour, and figurative usage;
• pronunciation variation between speakers and regions;
• translation ambiguity, where more than one rendering is defensible;
• grammar that depends on context rather than on a rule;
• natural speech variation, including what people say rather than what is taught.

Treat AI-generated language as educational guidance, not as a universal linguistic authority. Where a native or expert speaker in your specific context disagrees with the Platform, their context-specific knowledge deserves weight.

8. What "Mentor" Means

Mentor — the guiding instructional voice of the Platform, delivered through Educational Content and, where applicable, AI Features. A Mentor is not a certified teacher, examiner, translator, or adviser.

To be precise about what that means in practice:

• The Mentor is a designed instructional voice. It is not a named individual, and it is not represented as a person you are speaking with.
• Today the Mentor voice reaches you as authored Educational Content, reviewed by people before publication.
• Where AI Features are introduced, Mentor interactions may be generated by an AI Feature. In that case the Services will indicate that AI is involved.
• The Mentor is not always human, and no part of the Platform should be read as implying that a human instructor is reading your practice as you work.
• Where human instructors, tutoring, or human-reviewed feedback are offered in future, they will be described as such and distinguished from AI Features.
• Mentor guidance is educational guidance. It is not professional advice.

9. Human Oversight

Human oversight is a governance principle, and it is described here without overstatement.

What is accurate today:

• Educational Content is reviewed by people before it is published, including content drafted with AI assistance.
• AI systems, prompts, and configurations used in production are chosen, tested, and adjusted by people.
• Content may be corrected after publication when an error is identified, including errors a learner reports.

What this Policy does not claim:

• Xmentra does not claim that every AI-assisted output is individually reviewed by a person.
• Where live AI Features are introduced, individual responses generated for you in the moment will not, by their nature, be reviewed by a person before you see them.

Where live AI Features operate, oversight is exercised proportionately: systems may be evaluated before and after release, outputs may be sampled and assessed for quality and appropriateness, and safeguards may be applied and adjusted. Guaranteed individual human review is stated only where it actually exists.

10. AI-Assisted Authoring of Educational Content

Different kinds of Content carry different review processes, and it would be misleading to describe them as one.

• AI-assisted drafting. AI tools may be used while writing, structuring, or adapting Educational Content.
• Editorial review. Educational Content is reviewed by people before publication. Review is applied to published material as a whole rather than claimed as line-by-line verification of every possible reading of every sentence.
• Static published Educational Content. Lessons, learning experiences, explanations, and blog articles are fixed at publication, so an error can be found and corrected for everyone.
• Live AI-generated responses. Where such responses are introduced, they are produced at the moment of use and are not pre-published or pre-reviewed. They should be treated with more caution than published Educational Content.

Section 11 of the Disclaimer (/disclaimer) owns the description of accuracy limits in Educational Content generally.

11. Information You Send to AI Features

Where AI Features accept input from you, please send only what the learning task needs. Language practice rarely requires personal detail.

We ask you not to submit, unnecessarily:

• highly sensitive personal information, including health, biometric, religious, political, or sexual-life information;
• passwords, access codes, or security answers;
• payment card numbers, bank details, or other financial credentials;
• identification documents, or images of them;
• confidential information belonging to an employer, client, or other third party;
• information about another person that you do not have permission to share.

This section is about sensible input, not about processing. How Personal Information is collected, used, shared, retained, and deleted — including anything you send to AI Features — is owned by the Privacy Policy (/privacy).

12. AI Training

Information you submit through the Services, including Learning Data and text or audio you send to AI Features, is not used to train general-purpose AI models by default, whether by Xmentra or, to the extent our provider terms allow us to require it, by a Third Party Service. This matches the commitment in the Privacy Policy (/privacy), which owns it.

If this practice ever changes, the change will be described in the Privacy Policy and in this Policy before it takes effect, and where applicable law requires consent, consent will be requested rather than assumed.

Because these ideas are often blurred, they are kept separate here:

• General-purpose model training — using submitted information to improve a model that serves purposes beyond your learning session. This is not done by default.
• Service operation — processing needed to produce a response or run a feature at the time you use it.
• Safety evaluation — checking output or use for abuse, harm, or misuse.
• Quality evaluation — assessing whether a feature teaches well, including sampled review of outputs.
• Aggregate improvement — drawing conclusions from combined, non-identifying measures about how the Services are used.
• Platform analytics — aggregate measurement of Platform usage, owned by the Privacy Policy and described in the Cookie Policy (/cookie-policy).

Nothing in this section promises more than the Privacy Policy commitment, and nothing in it narrows that commitment.

13. Third-Party AI Providers

Where AI Features operate, they may rely on Third Party Services that provide model hosting, inference, speech synthesis, speech recognition, or related capability. Those providers are described here by role rather than by name.

Named providers are listed only in the Privacy Policy and, later, in a Subprocessors List. Where a provider processes Personal Information on our behalf, that processing is governed by the Privacy Policy (/privacy), which owns the subject.

Where a provider is used, we seek terms consistent with the commitments in this Policy, including the treatment of training data in section 12. We do not claim control over a provider's internal systems beyond what our agreements with them provide.

14. Personalisation

Where AI Features or the Services personalise what you see, personalisation may draw on information such as:

• your learning progress, including experiences and lessons completed;
• goals or preferences you have stated;
• previous practice, including items saved or revisited;
• confidence responses you have given;
• learner preferences such as pace or display settings.

Personalisation of this kind is educational sequencing: it affects what the Platform suggests you learn or revisit. It is not psychological profiling, it does not attempt to infer sensitive characteristics about you, and it is not used for high-stakes decisions about you. Today, Learning Data of this kind sits in Local Storage on your device, as described in the Privacy Policy.

15. Confidence Indicators

The Platform records how ready you feel about what you have practised, and uses those responses to suggest what to revisit. Confidence indicators are educational signals and nothing else.

They are not:

• a psychological diagnosis or assessment;
• a clinical or medical measure of any kind;
• a measure of intelligence or aptitude;
• an official proficiency level or certification;
• a prediction or guarantee of how you will perform in a real situation, including an examination.

Confidence responses are yours. A confidence signal is a prompt to practise again, not a judgement about you. This is consistent with the Disclaimer (/disclaimer), which owns the treatment of learning progress indicators.

16. Your Saved Practice Material

The Services keep a record of the language you have worked with, so that revision can build on what you have actually met rather than on a generic list. Its purpose is educational continuity: it lets the Platform bring back the right phrase at the right time.

Where you write or record something yourself, that material is yours. Being processed or displayed by the Platform does not transfer ownership of it to Xmentra. Ownership, licences, and permitted use are owned by the Terms of Use (/terms) and the Copyright & Intellectual Property Policy (/copyright-policy); this Policy does not create ownership rules.

Where this material is stored, how long it is kept, and how you can remove it are owned by the Privacy Policy (/privacy) and the Cookie Policy (/cookie-policy).

17. High-Stakes Use

AI-assisted educational guidance should not be your only source of truth where a decision carries real consequences. This includes:

• immigration and visa requirements;
• official examination requirements, formats, and results;
• university or school admission requirements;
• employment, professional registration, or credential requirements;
• legal matters;
• medical or health matters;
• financial matters;
• government forms, procedures, and deadlines.

For anything in that list, verify the current position with the relevant authoritative source — the examination provider, the authority, the institution, or a qualified adviser who knows your circumstances. This section describes where to check; it is not itself advice, and the Disclaimer (/disclaimer) owns the professional-advice boundary.

18. Examination Preparation

Where the Services help you prepare for a language examination, including with AI assistance:

• examination formats, requirements, and scoring are controlled by the examination provider, and may change without notice to us;
• AI-assisted practice, scores, or feedback are educational only and are not an official examination result;
• performance in practice does not guarantee performance in an official examination;
• you should confirm current requirements with the official examination provider before relying on them.

Xmentra is not affiliated with, endorsed by, or accredited by any examination organisation. Examination names, where they appear, are used descriptively to identify the examination being discussed.

19. Translation

AI-assisted translation is context-sensitive and imperfect. A rendering that is right in one setting may be wrong in another, and short phrases are especially prone to losing the meaning that context supplied.

Translations offered through the Services are educational. Where a translation matters — a document, a legal or medical text, an application, or anything official — have it produced or checked independently by a qualified translator. Xmentra does not provide certified or sworn translation.

20. Pronunciation and Audio

Audio on the Platform is provided to support listening and speaking practice. It may be recorded, synthesised, or delivered through other supported technologies.

Audio may vary by:

• accent;
• dialect;
• region;
• individual speaker;
• the synthesis or delivery technology used;
• the context in which a phrase is spoken.

Any single example represents one way of saying something, not the only correct way. Differences between the Platform's audio and a speaker you meet are usually variation rather than error.

21. Safety and Inappropriate Output

Where AI Features generate output, Xmentra applies reasonable safeguards intended to reduce inappropriate, unsafe, or harmful results. Safeguards may include configuration choices, restrictions on what a feature will do, monitoring, and evaluation.

Safeguards reduce risk; they do not eliminate it. No filtering is perfect, and we do not claim that inappropriate output can be prevented in every case. Language learning also involves difficult real-world subject matter, so context matters in judging what is appropriate.

If you encounter output that seems inappropriate, harmful, or seriously inaccurate, please tell us at hello@xmentra.com. Reports help us correct content and adjust safeguards.

22. Automated Decisions

Xmentra does not make decisions about you by automated means that have legal effect or a similarly significant effect on you. Nothing on the Platform is graded, certified, gated, or refused by a model.

Where AI Features or personalisation operate, their effect is limited to your learning experience: what is suggested, what is revisited, and how material is presented. If Xmentra ever introduces automated processing with significant effects, it will be reviewed under the maintenance architecture, described in this Policy and the Privacy Policy before it takes effect, and accompanied by the safeguards applicable law requires.

23. Younger Users

The Platform is intended for Users aged 14 or over, and is not directed at children under 14. This applies to AI Features in the same way as to the rest of the Services.

Where you are under the age of majority in your jurisdiction, you should use the Services, including AI Features, with the involvement of a parent, guardian, or teacher where that is expected of you locally. Requirements for parental consent differ between countries, and this Policy does not attempt to state a single rule for every jurisdiction.

Children's privacy, including how we respond if we learn that a younger child has used the Services, is owned by the Privacy Policy (/privacy). Eligibility is owned by the Terms of Use (/terms).

24. Your Responsibilities and Ours

Responsibility here is shared rather than transferred.

You are asked to:

• use your own judgement with AI-assisted output;
• verify important or high-stakes information with an authoritative source;
• avoid submitting sensitive or unnecessary personal information;
• avoid submitting information about others that you are not entitled to share;
• use AI Features in line with the Acceptable Use Policy (/acceptable-use);
• tell us when something looks wrong.

Xmentra takes on:

• describing AI involvement honestly, including its limits;
• reviewing Educational Content before publication;
• choosing, configuring, and testing AI systems with care;
• applying proportionate safeguards to reduce inappropriate output;
• correcting Educational Content when an error is identified;
• keeping this Policy, the Privacy Policy, and the Disclaimer aligned with what the Services actually do;
• not using submitted learning information to train general-purpose models by default, and disclosing any change before it takes effect.

Nothing in this Policy attempts to place the whole risk of AI-assisted learning on you. Liability is owned by the Terms of Use (/terms).

25. Changes to AI Features and to this Policy

AI functionality evolves, and this Policy is maintained alongside it rather than after it.

Changes affecting any of the following trigger a review of this Policy, the Privacy Policy, the Disclaimer, and the Terms of Use in the same release:

• how information is used by AI Features;
• model training practice;
• the role of a Third Party Service in AI processing;
• the introduction of automated processing with significant effects;
• safety measures and safeguards;
• personalisation and the information it draws on.

Alongside that:

• the Last Updated date changes with every edit;
• the Version number increases: a MAJOR increase for a change to rights, obligations, data practice, or scope; a MINOR increase for clarification, formatting, or contact updates;
• a MAJOR change is reflected here before or at the time the relevant feature becomes active, and, once Accounts exist, announced at least 14 days before its Effective Date;
• superseded versions are retained internally for reference;
• this Policy is reviewed at least every 12 months.

26. Contact Information

Questions about this Policy, or reports about AI-assisted output, can be sent to hello@xmentra.com.

27. Revision History