AI transparency

In plain language

Grace is an AI assistant, not a person or a qualified professional. It generates answers using your request, selected conversation history, instructions, and relevant knowledge sources. External AI services may process this information. Answers and citations can be wrong.

What AI processes

Depending on the feature, an AI request may contain your prompt, prior messages, relevant excerpts from family documents, images or other attachments, and context needed to respond appropriately. Document processing can extract text and create numerical representations for search.

Available features may generate text or images, interpret attachments, retrieve family knowledge, or use web search. When web search is used, search queries are sent to the search provider. Do not include unnecessary secrets or sensitive information.

Family knowledge and visibility

Family documents can help ground an answer, but retrieval does not establish that a source is correct, current, or complete. Review sources and important outputs yourself.

Parents can review child chats and manage child access. Joining a conversation makes the parent’s participation visible. Private adult and senior chats are not made available to parents simply because of their parent role. See children’s privacy.

Safety and limitations

Child messages are checked using automated safety processing. Safeguards can miss harmful content or block harmless requests. Grace is not an emergency service, and automated checks are not a promise that a person is watching every conversation.

Outputs can contain fabricated facts or citations, bias, or inappropriate suggestions. Do not rely on Grace alone for health, legal, financial, or safety decisions. Involve a qualified person when the stakes matter.

Providers and training

The provider used can depend on the model and feature configured for the family. See our service-provider disclosure for processing categories.

[The legal and operations owners must confirm provider-specific retention, training restrictions, opt-out settings, and any use of content for product improvement.] This draft does not promise zero retention or that every provider excludes every request from training.

Questions and feedback

Use Contact or the support channel in your workspace to report concerning output. Avoid sending sensitive conversation content until a suitable follow-up channel has been arranged.