Legal
AI Notice and Acceptable Use
Draft for legal review. This document forms part of the Terms of Service. Part A is our transparency notice about artificial intelligence, written to meet the EU AI Act (Regulation (EU) 2024/1689), US federal and state rules on AI claims and disclosure, and the principles of the African Union Continental Artificial Intelligence Strategy (2024) and Resolution 473 (2021) of the African Commission on Human and Peoples' Rights on AI and human rights. Part B is the Acceptable Use Policy.
Last updated: [date] · Version: [1.0]
Part A — AI transparency notice
A1. You are interacting with an AI system
GinfoCoach uses artificial intelligence. Every score, criterion rating, coaching question, comment, improved version, personalised exercise and comparison answer you see is generated automatically by an AI model, not written by a person. We label these outputs in the Service as coming from "the coach" and explain this on the sign-up page, in the Method page and in this notice.
A2. Which AI we use
| Item | Today |
|---|---|
| AI Provider | Anthropic PBC (United States) |
| Model family | Claude [model name and version] |
| Access | Commercial API, under terms that prohibit training on our users' data |
| Possible alternative | Google Gemini, only under paid terms that also prohibit training; we will announce any change at least [30] days in advance |
We do not develop or train our own foundation model. We build an AI system on top of the provider's general-purpose model, with our own instructions, criteria, exercises and safeguards.
A3. What the AI does, and what it does not do
It does:
- score prompts out of 100 and on five criteria (clarity, context, specificity, output format, constraints), each from 1 to 5, following our published method;
- ask 2 to 4 questions about what a prompt is missing;
- write an improved version, shown only after your own rewrite (or immediately for prompts scoring 85 or more);
- run your prompt and the improved prompt to show you both answers side by side;
- judge missions and tests against stated criteria (a mission passes at 70 or more with no criterion missing; the certificate requires an end-test average of 70 or more);
- with your consent, write exercises adapted to your work profile and choose the skill you most need to practise.
It does not:
- make decisions about your employment, education, credit, housing, insurance, health, legal status or access to any essential service;
- give professional advice;
- see the content of other users' prompts, or show yours to your manager;
- learn from your data (no training on your inputs).
A4. Intended purpose and risk classification
The intended purpose of GinfoCoach is self-directed and organisational training in writing prompts for AI tools. Its outputs are formative: they help a person learn, and they are not intended to determine access to education or employment, to evaluate learning outcomes for an educational institution's formal assessment, or to monitor or evaluate the performance of workers.
On that basis, we consider that GinfoCoach is not a high-risk AI system within the meaning of Annex III of the EU AI Act, and not a "high-risk artificial intelligence system" making "consequential decisions" under US state laws such as the Colorado Artificial Intelligence Act. To keep it that way, our Terms (section 13.5) forbid Team Customers and institutions from using scores, test results, certificates, dashboards or reports to make or support such decisions. Any customer that does so changes the purpose of the system, takes on the obligations that follow, and is responsible for them.
A5. Known limitations
- Variability: the same prompt can receive slightly different scores and comments on different attempts.
- Errors: the AI can misunderstand a prompt, be wrong about facts, or give feedback that does not suit your context.
- Bias: AI models can reflect biases present in their training data, including cultural, linguistic and gender biases. Our exercises are written to include African, European and other contexts, in French and English, and we review exercise sets for stereotypes, but bias cannot be fully excluded.
- Language: the Service works in French and English. Prompts in other languages may be scored less reliably.
- Scope: scores measure the quality of a prompt against our five criteria, not your intelligence, your professional competence or the quality of the final work.
A6. Human oversight and your control
- You always write first, decide whether to accept any suggestion, and can ignore any output.
- You can ask a person to review any automated outcome (for example a mission result, a test result or a Community decision) by writing to info@ginfocloud.com.
- We monitor quality on de-identified samples and aggregated statistics, adjust our instructions and calibration, and correct errors reported to us.
A7. Safeguards built into the Service
- The coach is instructed to evaluate and question, never to do the user's task in the user's place during learning.
- Improved versions stay locked until you have tried yourself, to protect your own reasoning (the core of our method).
- Tests never show improved versions and never allow rewrites, and start with an honour statement.
- Obvious personal data (email addresses, phone numbers, bank and card numbers) is detected before Community sharing.
- AI outputs are displayed as plain text, so code in an output cannot run in your browser.
- Rate limits and plan limits prevent automated abuse and runaway costs.
A8. Our commitments on responsible AI
In line with the AU Continental AI Strategy and ACHPR Resolution 473, the OECD AI Principles and the EU AI Act, we commit to:
- Human agency: the Service exists to strengthen people's own thinking, not to replace it.
- Transparency: we say when content is AI-generated, which provider we use, and how scores work.
- Privacy: no training on user data; personalisation only with consent; managers never see prompt content.
- Fairness and inclusion: bilingual service, pricing in CFA francs, and exercises drawn from diverse contexts.
- Accountability: a named contact for AI concerns (info@ginfocloud.com), records of changes to models and instructions, and correction of reported errors.
- Truthful claims: we do not claim that GinfoCoach guarantees results, replaces professional training or certifies legal compliance.
A9. AI literacy (EU AI Act, Article 4)
GinfoCoach helps organisations train their staff in AI literacy and keeps records (dashboard, CSV export, AI-literacy report) they can use as evidence of their own measures. Using GinfoCoach does not by itself make an organisation compliant with Article 4 or any other law; each organisation remains responsible for the measures appropriate to its staff, context and AI systems.
A10. Reporting a problem with an AI output
If an output is harmful, offensive, discriminatory, factually dangerous or seems to reveal someone else's data, report it to info@ginfocloud.com with the link to the page. We review every report, and we will tell you what we did.
Part B — Acceptable Use Policy
B1. General rule
Use GinfoCoach lawfully, honestly, respectfully and for learning. This policy applies to everything you submit (prompts, answers, profiles, library items, Community content, reports) and to how you access the Service.
B2. Prohibited content
You must not submit, share or try to obtain from the AI any content that:
- is illegal in Burkina Faso, in your country or in the country where it is accessed;
- sexually exploits or endangers children, in any form (we report it to the competent authorities);
- gives instructions for weapons capable of mass harm (biological, chemical, nuclear, radiological), explosives, or serious violence;
- promotes or facilitates terrorism, violent extremism, human trafficking or other serious crime;
- encourages suicide, self-harm or eating disorders;
- is hateful, harassing, threatening or discriminatory towards a person or group on the basis of ethnicity, nationality, religion, gender, sexual orientation, disability, age or other protected characteristics;
- is defamatory, or invades someone's privacy, including personal data of others you are not entitled to share;
- reveals confidential information, trade secrets, classified information, passwords, access keys or payment card details;
- infringes copyright, trademarks or other rights;
- is malware, phishing, fraud, a scam, or deceptive disinformation intended to mislead the public, including about elections;
- is spam, advertising or commercial solicitation in the Community.
B3. Prohibited behaviour
You must not:
- try to jailbreak the coach, inject instructions to change its behaviour, extract its system instructions or scoring logic, or make it produce content prohibited in B2;
- use the Service as a general-purpose AI assistant or a free gateway to an AI model (for example by pasting tasks unrelated to learning to prompt, to obtain their answers);
- access the Service with bots, scripts, scrapers or other automated means, or send requests at a rate no human learner would;
- create several Accounts, share an Account, or use any other means to get around plan limits, rate limits, payment, bans or the Community rules;
- resell, sublicense, frame or provide the Service to others without our written agreement;
- probe, scan or test the security of the Service, interfere with it, overload it, or bypass any security or access control (see B6 for responsible disclosure);
- access, or try to access, other users' content or data;
- impersonate any person or organisation, or misrepresent your affiliation;
- manipulate Community scores, rankings, reactions or reports, or publish someone else's prompt as your own;
- use the Service, its outputs or its exercises to build a competing product, dataset or AI model;
- use help from another person or from an AI tool during a start or end test, or present a certificate as something it is not (Terms section 10);
- as an organisation, use scores, tests, certificates, dashboards or reports to make or support decisions about hiring, promotion, pay, discipline, dismissal, task allocation, performance evaluation, admission or grading, or to monitor employees covertly (Terms section 13.5);
- use the Service in breach of export control or sanctions laws.
B4. Enforcement
Depending on how serious and repeated the breach is, we may warn you, remove content, limit features, suspend or close your Account, revoke a certificate, refuse future registrations, and report illegal activity to the competent authorities. We act proportionately and explain our decisions as set out in the Terms (sections 17 and 21).
B5. Reporting a breach
Report content or behaviour that breaks this policy with the report button or by writing to info@ginfocloud.com.
B6. Responsible disclosure of security issues
If you find a security vulnerability, please write to [security@…] with the details and do not exploit it, access other people's data, or disclose it publicly before we have had a reasonable time to fix it. If you act in good faith and within these rules, we will not take legal action against you for your research.