AI & Content Transparency

In short

BrainSpark uses AI to create fresh educational questions. Some selected topics use curated reference material. Questions are shown immediately, and most automated quality review happens afterward. BrainSpark is designed to support learning, but AI-generated content can contain errors.

AI-assisted generation

BrainSpark uses Anthropic's commercial AI services to generate questions, answer choices, correct answers, explanations, rewrites, and certain quality-review outputs. Generation is based on the selected age band, learning region, topic, subtopic, difficulty, and limited recent-question context used to reduce repetition.

What “grounded” means

For selected Indian history topics covering the Maurya, Gupta, Chola, Vijayanagara, and Maratha periods, BrainSpark uses curated source material developed with reference to relevant official NCERT textbook chapters. This grounding reduces reliance on general model knowledge, but it does not guarantee that every generated statement is correct or directly traceable to a source.

What is not grounded

Other India content, U.S. content, STEM topics, and general topics may rely on prompts, prior question context, and the AI model's capabilities without the same NCERT-based source package. BrainSpark does not claim that all questions, all Indian history content, or all curriculum content is NCERT-based or independently fact-checked.

How review works

1. A question request is created from the learner's selected context.
2. The question is generated and shown to the learner immediately.
3. The generated question and related quality metadata are stored in the question bank.
4. For most non-grounded topics, an automated review runs after display and records a validation result or status.
5. For the selected grounded Indian history topics, the separate post-generation validator is skipped because curated reference material is already used during generation.
6. A user report may later lead to human review, correction, retirement, or deletion.

Important: A learner may see and answer a question before automated validation is complete. A validation status does not prove accuracy and does not mean that a human approved the question.

What the automated review does

• Checks for possible internal inconsistency.
• Looks for certain recurring misconceptions or unsupported claims.
• Compares against prior question-bank content to identify duplication or contradiction.
• Checks certain age- and content-appropriateness concerns.
• Adds quality metadata that can support later review and question-bank improvement.

What the automated review does not do

• Guarantee correctness, completeness, neutrality, age calibration, or curriculum alignment.
• Automatically block every uncertain or potentially inaccurate question before display.
• Replace teachers, parents, textbooks, professional advice, or human judgment.
• Guarantee that every flagged question will receive human review within a particular timeframe.

Human review and reporting

The normal BrainSpark interface provides fixed-choice report reasons rather than free-text reporting. A report is a signal for review, not a promise of a particular result or response time. BrainSpark does not currently operate a routine deletion schedule for generated question records.

Data sent to AI providers

Educational prompt and question context sent to Anthropic may include the selected learning region, age band, topic, subtopic, difficulty, recent question text, and generated educational content. BrainSpark does not intentionally include a learner's name, email address, account identifier, device identifier, or IP address in the generation prompt.

According to Anthropic's current commercial-data policies, commercial API inputs and outputs are not used to train Anthropic models by default. Anthropic's applicable terms, retention practices, safety exceptions, and account settings govern how provider-side data is handled.

BrainSpark also uses Voyage AI to create vector embeddings of generated question text for semantic similarity and duplicate detection. The text sent for embedding does not intentionally include a learner's name, email address, account identifier, device identifier, or IP address. BrainSpark has opted out of Voyage AI's use of submitted content for model training. Voyage AI's applicable terms and retention practices govern provider-side handling.

Responsible-use guidance

• Verify important facts with reliable sources.
• Report unclear, inaccurate, or unsafe content.
• Do not use BrainSpark as medical, legal, financial, safety-critical, or other professional advice.
• Do not submit confidential or personal information.
• Adults should supervise younger learners and help interpret uncertain content.

Questions about AI or content quality?

Contact BrainSpark at support@quintelx.ai
Operator: QuintelX LLC
Website: brainspark.study