AI Policy

01

AI-Assisted Design Disclosure

Inflict.io recognizes that the TTRPG publishing community is filled with tremendous artists and creative minds. We support independent RPG creators and their work to the best of our ability and always will.

That said, Inflict.io is a very small team. We are not artists or graphic designers and have no training in these disciplines. We launched this project as a part-time hobby, working without any real budget. For these reasons, while the text and ideas found in our modules and scenarios are entirely our own, certain graphical design elements and artwork were created using generative AI. We are aware that the use of generative AI in creative projects is controversial, but without it, this particular project could not exist.

It is our hope that should our material reach an audience who loves it and shares the news about us, our little crew will grow, and so will our art budget. Should that occur, Inflict.io would look forward to incorporating the contributions of the many wonderfully talented human artists who deserve to have their work seen.

02

Questionnaire Quick Reference

Survivor and Unlimited hosts can use OpenAI to fill a character Quick Reference once. The host must select Fill once with AI and confirm the request. If notes exist, Inflict.io warns that the AI fill will replace them. The AI fill replaces those notes only after the host confirms.

Inflict.io sends only submitted question-and-answer pairs and a one-way account pseudonym to OpenAI. We do not send character descriptions, host notes, player email addresses, account IDs, invitations, or current answer drafts. The API request sets store to false.

Our instructions require short notes that use only facts that the player wrote in submitted answers. Each AI note includes the numbers of its source questions. Unsupported categories stay empty. AI can still make an error. Hosts must compare each note with its cited answer before use.

Inflict.io stores the Quick Reference with the questionnaire. The host can edit the notes after the AI fill. The host cannot request a second AI fill for that questionnaire. Only the host and restricted administrators can read these notes. Players do not receive them. See our Privacy Policy for retention details.

03

Questionnaire AI review

Each host can enable Discourage AI answers for a scenario. The setting is off by default. When it is off, players can paste answers and we do not calculate or show an AI score. When it is on, the questionnaire warns players and blocks paste in their answers. A host view automatically requests a review after the questionnaire is complete. Hosts can also select Check full questionnaire to review available answers or retry a request.

The review considers all submitted answers together. Inflict.io sends their questions and answers, the character title and description, and a one-way account pseudonym to OpenAI. We exclude host notes, email addresses, account IDs, invitations, and saved drafts. The API request sets store to false.

The score estimates AI assistance anywhere in the submitted answers. It is an uncalibrated model estimate, not a measured probability or the percentage of AI-written text. The model can miss AI use or flag human writing. Questionnaires with fewer than 300 answer words receive an inconclusive result without an OpenAI request. The model can also return an inconclusive result when evidence is insufficient.

Only the host and restricted administrators can read the saved result and its cited evidence. Changes to answers, questions, character context, or the assignment invalidate the result. The review does not reject answers or restrict players.

04

Contact Inflict.io LLC

Questions about this disclosure can be sent to support@inflict.io.