Archivist logo

FAQ's

Data Privacy and Security

We only store two types of data for your campaigns: text and images. Text is stored as a chronicle of what happened during the session, either as a speech-to-text transcript or plain text ingested through Discord channel message history, plain text uploads, Roll20 chat archives, etc. If any audio is used to generate a session chronicle, Archivist will not store any of that data. Uploaded images are stored for use as avatars/tokens for Character, Faction, and Location objects associated with your campaign.

The user has full visibility into all stored text through a transcript download on the details page of each Session. The user can invite other users (players) to their campaign, allowing them access. The user has full discretion to remove player access at any time. Currently, the internal Archivist team may also access user data for the explicit purpose of improving the product. This access will likely be greatly curtailed, anonymized, or revoked entirely once Archivist is fully launched.

Yes. Any content (text or images) provided by the user can be permanently deleted by the user at any time. Users may also permanently delete any content (summaries, moments, etc.) generated by Archivist at their discretion.

No. Archivist does not share any user data with third parties for advertising or any other purpose. Any data collected on the platform is used solely for the purpose of improving the product.

The World Chatbot uses a technique called Retrieval Augmented Generation (RAG). This means Archivist creates embeddings of campaign session chronicles so when the user asks a question, Archivist can quickly and effectively retrieve the information relevant to that question's answer. Embeddings are also permanently deleted when a Session is deleted.

We take several measures to ensure your data is secure:
  • Encrypted Communication: All data is encrypted both in transit and at rest using industry-standard protocols like HTTPS and SSL/TLS.
  • Secure Authentication: We implement token-based authentication with support for multi-factor authentication (MFA) to ensure secure user access.
  • Protected Infrastructure: Our platform is hosted on secure infrastructure with built-in DDoS protection, automated backups, and role-based access controls to minimize attack vectors.
  • DNS Security: We use DNS management services with additional protection against DDoS attacks and other common vulnerabilities.
  • Isolated Environments: Our backend services and integrations run in isolated environments to further safeguard your data.
Features and Functionality

Yes. Users have full control over which data Archivist is exposed to. Our speech-to-text feature has explicit controls for the user to easily start and stop the speech-to-text service, and Archivist will never ingest or transcribe audio without the authorized user’s consent.

Speech-to-text is not perfect and is especially challenging when transcribing fantasy-based lexicons as much of that vocabulary is invented. However, Archivist is optimized specifically for your campaign setting to ensure the highest level of accuracy possible.

If Archivist experiences connectivity issues or service is interrupted for any other reason before a Session has ended, the state of the transcript is saved. This means any transcribed text up until the point of service interruption is preserved, and the user can restart the speech-to-text service, picking up where it left off.

No. The World Chatbot is also available directly in your Discord server for the rest of your player group to access.
General Questions

All of a user’s account data will be preserved upon cancellation in case they decide to resubscribe. Full account deletion is available upon request by emailing [email protected]

Currently, users can export Session transcripts directly. The Archivist team is working to add an option for a full account data archive download in the future.

At Archivist, using AI ethically is one of our core principles. One of the challenges we face is how to support users whose creative strengths may not lie in visual arts while also recognizing and fairly compensating the artists whose incredible work has made image generation models possible. We’re currently collaborating with companies that provide models trained on assets volunteered by artists in exchange for fair compensation. Our goal is to help create a scalable revenue stream for these artists—something that wasn’t widely available before AI image models became popular. In the meantime, for product development purposes, we are using off-the-shelf models trained on publicly available images that don’t offer direct compensation to artists. While this is necessary at this stage, we are committed to moving away from these models as soon as possible.