Alice

Privacy and data

What Alice protects, what it does not, and what happens to your data. This describes the software as built; it is not a certification.

What is protected, and how

Every read and every change is checked on the server against your member identity: your conversations, memories and briefs can be read only by you. Pages never receive someone else's private material to hide it in the browser.

Matching runs on the server and reads only brief versions you authorized for matching, in communities you chose. It never reads conversations or memories.

Anything shown to another member is an exact text you approved, for that person and that opportunity, and only after both of you approved. Approvals are recorded per person and bound to a hash of the exact text; an edit needs a new approval, and every condition is checked again at the moment of delivery.

Community administrators manage invitations and membership. They cannot read anyone's conversations, briefs, memories or opportunities.

What this is not

It is not end-to-end encrypted. When live AI is configured, the text of your conversation with Alice is processed in plain text on our servers and by the AI provider to produce her replies. Data is encrypted in transit; at rest it is protected by the database provider's storage encryption.

It is not legal confidentiality. Do not share material you are not entitled to share, and do not rely on Alice to protect trade secrets or settle ownership, authorship or payment.

AI processing

Live replies use Anthropic's Claude through its API, when configured. A model call receives only your own conversation, the memories you confirmed and your own briefs. Another member's private material is never sent in a call made for you. Text from members is passed to the model as material, not as instructions, and the model cannot change permissions, choose recipients or disclose anything: those are deterministic server rules.

Without AI credentials, the demo uses a scripted Alice, labeled on every reply. Your content is not used to train models or for unrelated analytics. The AI provider's own retention terms apply to what it processes; check them before launch.

Memory

Alice may suggest remembering something you said about yourself. Suggestions are not used until you confirm them. You can see where each memory came from, correct it, exclude it from brief drafts, delete it, or turn memory off. "Do not remember this" on a message stops any memory from coming from it and removes any already suggested.

Withdrawal, deletion and export

Withdrawing an approval or blocking someone stops future access to what you shared, where Alice controls it. It cannot erase what the other person has already read or copied.

Export gives you your own content and what others chose to share with you, as one JSON file. It never includes Alice's internal matching notes or anyone else's private material.

Deleting your account removes your profile, conversations, memories, briefs and their versions, matching permissions, the payloads you shared, pending opportunities and your messages in introductions. The other person in an introduction keeps their own messages. Database backups may retain deleted data until they expire under the database provider's backup schedule; it is not restored except to recover from a failure.

Notifications and analytics

Notices say only that something needs your attention, never what a message, brief or match says. External notices are off by default and are not sent in the demo. Logs and analytics exclude conversation text and brief content; aggregate counts of AI use are kept to judge costs.

Billing

Payments are handled by Stripe. Card details go to Stripe's hosted pages and never to Alice; Stripe receives your member id, your email if you give it at checkout, and the plan, never conversation text, briefs or match reasons. Alice's billing records hold your plan and usage counts only.