Jumbly legal
AI Policy
A transparent overview of Jumbly's AI-assisted puzzle direction, review expectations, safety standards, and current beta limitations.
Last updated: May 11, 2026
These pages describe the current Jumbly closed beta and planned platform features in plain language. We will keep them updated as accounts, analytics, subscriptions, and wider launch systems mature.
How Jumbly May Use AI
- AI may assist with puzzle candidate generation, hints, explanations, categories, quality checks, ambiguity warnings, and admin review workflows.
- This may include emoji-based puzzle ideas, emoji meaning slots, phrase clues, and explanation drafts for games such as EmojiWords and Emojilock.
- AI generation should be used as an editorial tool, not as an automatic publishing system.
- Normal gameplay should remain playable without sending player guesses to an AI provider.
Human Review Direction
- Generated puzzle candidates should be reviewed for fairness, clarity, offensive content, ambiguity, answer quality, and cultural sensitivity before publication.
- Jumbly may edit, reject, approve, remove, regenerate, or republish AI-assisted puzzle content.
- Admin review and approval workflows should be protected before any real AI generation is exposed publicly.
Quality and Fairness
- AI-assisted puzzles can be wrong, too ambiguous, too obscure, repetitive, culturally narrow, or unintentionally inappropriate.
- Jumbly should use validation checks, duplicate detection, difficulty scoring, fairness analytics, player feedback, and moderation warnings to improve puzzle quality.
- A puzzle being AI-assisted does not guarantee it is correct, fair, or suitable for every player.
Content Standards
- AI-assisted content should avoid hateful, harassing, sexual, exploitative, violent, illegal, misleading, or otherwise harmful material.
- Puzzle content should avoid unnecessary use of living people, sensitive personal traits, private information, and trademark-sensitive or copyrighted outputs where they create legal or quality risk.
- Jumbly should keep a review path for player reports about unfair, offensive, confusing, or low-quality AI-assisted puzzles.
Provider Transparency
- If OpenAI or another provider is enabled, Jumbly should document the provider, purpose, data sent, retention expectations, and safety checks.
- API keys must remain server-only and must never be exposed to the client.
- If AI generation is unavailable, Jumbly should fall back to human-authored or device-saved candidate data without crashing.
Current Status
Jumbly currently treats AI puzzle generation as infrastructure and governance work. Real production use should wait until admin permissions, validation, moderation, provider configuration, and review workflows are hardened.