AI Ethics

Blabble exists to help students learn the material, not to launder assignments. “Built to teach, never to cheat” is the test every feature has to pass.

A study tool that invents confident nonsense is worse than useless. Blabble’s AI is built to work from your materials.

• Quizzes, flashcards, mind maps, plans and practice exams are grounded in the documents you upload, so they reflect your actual syllabus.
• We’re honest that AI can still be wrong, and we tell you to verify — grounding lowers the risk of hallucination, it doesn’t remove it.
• We don’t dress up guesses as facts; outputs point back to your own content.

The student is the protagonist, not the AI. Every AI output can be regenerated, edited, ignored or deleted, and a human always interprets it. Automated scores and readiness signals are framed as guidance to act on, never as judgements handed down. Real people are reachable for any AI output you want to question.

• You’re always told when you’re interacting with AI, and AI-generated material is labelled.
• We document the model roles, routing safeguards and what we send — see the AI Use Policy. Specific models can change as quality and safety improve.
• We state the limits plainly: AI can be wrong, grades are indicative, and you should check before you rely.
• We don’t overclaim. Blabble is a study companion, not an oracle or a teacher-replacement.

The detail lives in our AI Use Policy.

AI models carry the biases of their training data. We can’t fully solve that, but we take it seriously and design to reduce harm:

• Grounding in your own materials keeps output anchored to your curriculum rather than a model’s assumptions.
• We focus on the IB context we understand well, rather than over-claiming universal expertise.
• We treat AI grades and feedback as coaching to be checked, precisely because automated assessment can be uneven.
• When we find the AI being unfair, unhelpful or wrong for particular students, we treat it as a bug to fix — please report it.

Ethical AI starts with not exploiting the people who use it.

• Your core study data lives on our application server within the European Economic Area.
• We send the AI only what a task needs, not your whole account.
• AI requests use zero-data-retention routing and deny providers that collect prompt data.
• We do not train our models on your work, and we never sell your data or run advertising.
• You can request a copy of your data or delete your account.

The full picture is in our Privacy Policy.

Many of our users are minors, so safety isn’t optional. Blabble screens AI chat with an independent moderation model, blocks clearly harmful content, and handles sensitive academic topics with extra care rather than refusing students who legitimately need to study them. When a safety check is uncertain, we err toward protecting the student.

Principles mean nothing without a way to challenge us. If an AI output is wrong, unfair or unsafe, you can tell us and a human will look at it. We fix safeguards based on real reports, and we’re accountable for the AI we ship — we don’t hide behind “the model did it”.

These principles aren’t only values — they map onto our legal obligations under the GDPR and the EU AI Act. We classify Blabble as a limited-risk, transparency-focused educational tool, we keep human oversight in the loop, and we review our AI use as our features and the rules evolve. Where we fall short, we’d rather hear it and fix it than defend it.