AI-chat terms and boundaries

A fictional AI experience

Every crew participant is a fictional AI character simulation. Replies are generated by software, not typed by a live human, the author, or a person who may have inspired a character. The synthetic roster, typing indicators, availability labels, and scripted introduction do not mean a real person is present.

Generated replies are non-canon. They do not add to or change the books and must not be treated as an author statement, endorsement, or fact about a real person.

Accuracy and safety limits

AI replies can be wrong, incomplete, out of character, offensive, unsafe, or unexpectedly revealing despite input checks, guardrails, output validation, and reports. Moderation and personal-data detection are not perfect.

The chat is not a professional, emergency, medical, legal, financial, or crisis service. Do not rely on a reply for an important decision. High-risk situations may receive neutral system guidance outside character voice, but help is not guaranteed.

What you may send

• A visitor message must contain 1-600 Unicode characters after normalization.
• A completed exchange contains 1-3 AI replies, each no longer than 400 characters and no longer than 900 characters combined.
• Do not submit passwords, credentials, private records, payment data, contact details, identifying information, confidential material, or content you do not have permission to provide.
• Do not use the chat for operational intrusion, phishing, malware, evasion, credential theft, stalking, doxxing, target-specific attacks, safety-control bypass, corpus extraction, or attempts to impersonate a crew/system identity.

Blocked or failed input is not saved as transcript content. Abuse-rate and safe request-state metadata may still be recorded without the rejected text.

Private session, not confidential advice

The session is isolated from other visitors, but it is not private from the service and its listed Cloudflare processors. A host-only HttpOnly cookie identifies the session. The service processes the handle and messages to retrieve approved Book 1 context, generate replies, enforce policy, and resume the short-lived transcript.

Completed session data expires after 7 days of inactivity and no later than 30 days after creation. The transcript is capped at 120 messages. Clear session invalidates active access and removes active rows, subject to the plan-dependent D1 recovery caveat and the separately retained report exception described in the privacy notice.

Visitor messages are not used to train or fine-tune the characters and are not added to the Book 1 corpus. Reporting an AI message submits only a fixed reason and a minimal redacted generated excerpt for up to 30 days; there is no free-text report.

Relationship and presence boundaries

The crew must not claim consciousness, live employment, real-world observation, private device/account access, or a need for the visitor. Romantic or sexual roleplay, exclusivity, dependency, paid intimacy, guilt for leaving, streaks, and unsolicited re-engagement are outside this experience.

Visitors must confirm they are at least 13 before session creation. No exact birthdate is requested. A parent or guardian should supervise a minor's use where appropriate.

Availability and launch controls

Replies, particular characters, or the entire chat may be disabled for safety, privacy, budget, rights, maintenance, or provider reasons. The service does not promise uninterrupted access, a response to every message, or preservation of a session beyond the disclosed limits.

The owner/author has approved the fictional character simulation, Book 1 retrieval corpus, scripted introduction, public copy, 13+ posture, and selected Cloudflare-hosted model use for this product. That approval is owner risk acceptance, not an independent legal, rights-holder, or accessibility certification. Automated and live review can reduce risk but cannot guarantee every reply.