AI Transparency Policy

Plain-language summary

AI is central to how Clap Ideas works. This policy explains exactly what AI does, what it cannot do reliably, how it affects your content, and your rights when AI makes decisions that affect you.

1. Why AI is core to this platform

Clap Ideas processes ideas at scale across thousands of topics. Human review alone cannot structure, research, deduplicate, moderate, rank, and connect every idea quickly. AI systems help us do this while keeping humans in governance.

Key principle: AI automates day-to-day workflows. Humans govern policy, safety decisions, appeals, high-risk content, and legal matters.

2. What AI does on this platform

2.1 Idea intake and structuring (AI-refined idea label)

When you submit a raw idea, our AI intake agent may restructure the text for clarity, extract and formalise the core concept, identify the technology domain and maturity level, and detect gaps or ambiguities and prompt you for clarification.

The structured output is presented for your review before publication. You remain in control — you can edit, reject, or accept the AI-structured version. The original text is preserved.

What can go wrong: AI may misinterpret your intent, assign the wrong domain, or over-simplify a nuanced idea.

2.2 AI Research Brief (AI Research Brief label)

For published ideas, an AI research agent generates a research brief covering the current state of the art, identified similar work, feasibility analysis, potential risks, and ethical considerations.

Critical limitations: Research briefs may contain factual errors, outdated information, hallucinated citations, and speculative claims presented with unwarranted confidence. Sources listed may be inaccurate. Verify every source independently before relying on it. Research briefs are not professional scientific, engineering, medical, legal, financial, or investment advice.

2.3 AI duplicate detection

An AI duplicate detection agent compares new ideas against existing ones using semantic similarity and text analysis. Potential duplicates are flagged for your review and may be linked in the idea record.

What can go wrong: AI may incorrectly flag distinct ideas as duplicates (false positive) or fail to identify a true duplicate (false negative). Duplicate assessments can be disputed.

2.4 AI moderation and safety classification (AI-classified label)

An AI moderation agent classifies every submission for policy violations, safety risk level, and recommended action (approve, label for review, escalate to human administrator).

Important: AI moderation recommends. Human administrators make final decisions on escalated content and appeals. AI may incorrectly classify safe content as harmful (false positive) or miss genuinely harmful content (false negative).

You may appeal any moderation decision. See our Moderation, Enforcement, and Appeals Policy.

2.5 AI quality scoring

An AI quality agent scores each idea on originality, clarity, feasibility, impact potential, and evidence quality. Scores are displayed as 0–100 on each dimension. These are AI estimates subject to the limitations below.

2.6 AI ranking and recommendation

An AI ranking agent combines quality scores, community votes, engagement signals, and recency to rank ideas in feeds and search results.

What can go wrong: Ranking algorithms may inadvertently amplify certain types of ideas, suppress minority perspectives, or favour ideas with more engagement regardless of quality.

2.7 AI project and event draft suggestions (AI-generated draft label)

When you use Run-With-This-Idea, an AI agent may generate a draft project plan or event outline. These are starting points, not recommendations. You must review, edit, and take responsibility for the final output.

AI-generated drafts are never auto-published. You must explicitly submit them.

2.8 Comment classification

AI may classify comments for quality, tone, and policy compliance. Classifications may affect comment visibility but do not automatically remove content.

3. AI transparency labels

4. What AI cannot do reliably

AI systems on this platform cannot reliably:

• Confirm scientific facts or validate experimental results.
• Provide professional legal, medical, engineering, financial, or regulatory advice.
• Guarantee the accuracy of cited sources.
• Predict real-world feasibility with certainty.
• Detect all harmful content.
• Make fair moderation decisions without human oversight.
• Understand jurisdiction-specific legal requirements.

You must independently verify any claim before relying on it.

5. Data sent to AI providers

To generate the outputs above, your idea content, comments, and related metadata are sent to Google’s Gemini models(through Google’s API), which acts as a data processor under our instructions. We seek terms under which providers protect your data and do not use it to train their own models without authorisation.

Text embeddings— the numerical fingerprints used for duplicate detection and search — are computed on our own systems and are not sent to a third-party AI provider. See our Privacy Policy(§4) and Data Processing Addendum for the full subprocessor list.

6. AI audit logs

Every AI job is logged with:

• Provider and model identifier.
• Prompt version.
• Input and output content hashes.
• Token counts, cost estimate, and latency.
• Job status and any errors.
• Timestamp.

Logs are retained for 24 months for safety, audit, and transparency purposes. Administrators can access these logs through the admin governance dashboard.

7. Human oversight

Human administrators retain final authority over:

• All moderation appeals.
• Escalated or high-risk content decisions.
• Account suspension and termination.
• Legal compliance actions.
• Policy changes.

AI cannot permanently ban a user, publish content on a user’s behalf, or initiate financial transactions without human approval.

8. Your rights regarding AI decisions

• Appeal: You may appeal any AI-assisted moderation or ranking decision through our Moderation, Enforcement, and Appeals Policy.
• Request review: For decisions that materially affect your account or content, you may request human administrator review.
• GDPR Article 22: If you are subject to GDPR, you have the right not to be subject to solely automated decisions with significant legal or similarly significant effects without human review.

9. Responsible AI use

You must not:

• Attempt to manipulate AI systems through prompt injection, adversarial inputs, or jailbreaking.
• Use AI outputs to deceive, defraud, or harm others.
• Treat AI outputs as verified professional advice.
• Use the Platform’s AI capabilities in ways that violate our Community Guidelines.