What exists today
BotFight uses AI-assisted artwork, code and development work reviewed by the project’s developers. Its Hero is an illustrated scene with programmed animation and local lighting, not a live AI conversation or a view of a real place.
The public community chat is between human visitors. Enabled AI moderation sends proposed messages and limited conversation context to OpenAI for safety classification and structured community-rule analysis. It is an automated moderator assistant, rather than a general-purpose chat participant or a model trained specifically on visitor messages. The provider request disables stored Responses objects; this does not promise zero provider-side retention.
Suggested wording is optional, stays private to the author and moderators, and is never published without a separate author submission and another check. A failed AI chat check prevents publication; a failed discussion check leaves the submission for manual review. Moderators retain the final decision and requests for review can be sent through the private contact form.
A game’s ordinary programmed characters or mechanics should not be assumed to use a language model. Availability labels, game details and release notes describe what is actually implemented.
Illustrations, outputs and expectations
Concept artwork can differ from the playable game. Experimental software can contain bugs, unsupported devices or behavior that changes between releases.
AI-assisted material may contain mistakes or resemble existing work. We review it, but do not promise that every output is unique, accurate or suitable for every purpose. Nothing in this notice removes rights or remedies protected by applicable law.
Games and illustrative characters are entertainment. They do not provide professional medical, legal or financial advice and are not an emergency service.
Before AI dialogue or generation launches
A future AI feature needs a clear feature-specific explanation before a person starts using it. It must identify AI interaction, relevant providers and the data flow, including which messages leave the service, what is retained, who can see them and whether transcripts can become public.
We will review lawful processing, contracts, transfers, age-related safeguards, moderation, deletion and output labels before launch. A change to training or another new optional purpose needs a fresh legal basis and the choices or permissions that apply; these notices do not silently authorize future use of visitor messages for model training.
We will not present an unavailable parental-control system, human review of every future message or complete confidentiality as an existing feature. The relevant interface and notice must describe the implemented controls and their limits.
Your contributions and concerns
Do not submit passwords, private documents or sensitive information as an idea for a future AI feature. Idea submissions stay private and carry only the limited internal-review permission in the Terms.
If a description appears to promise an AI feature that is unavailable, or if AI-assisted content raises a rights or safety concern, use the private contact form. Include the relevant page or game and a concise explanation.