隱私政策

Novilot 是面向小說作者的寫作應用。手稿是你可以放入軟體的最私密內容，本頁會準確說明手稿和其他數據會發生什麼。簡而言之：寫作屬於你，我們只用它運行你請求的功能，絕不出售數據，也絕不使用你的文字訓練 AI 模型。

我們收集什麼

• 賬戶數據。 您的電子郵件地址和密碼的哈希版本（使用 bcrypt 進行哈希處理；我們看不到密碼本身）。
• 你的寫作。 項目、章節、故事聖經條目、大綱、筆記、章節版本，以及你在應用中創建的其他內容。這些內容會被保存，以便應用再次展示給你。除非你明確分享（例如創建 beta 讀者鏈接），否則它們不會公開。
• 人工智能工作數據。 To power features like autocomplete, continuity checking, and voice matching, we store material derived from your manuscript: summaries, style profiles, continuity observations, and text embeddings. These are private working data, subject to the deletion and retention details below.
• 使用數據。 你使用的功能和基本產品事件（例如運行了一次分析），以及用於公平計費的積分使用情況。我們還會統計你為個人目標和連續寫作記錄寫下的字數。
• 計費數據。 Payments are handled by Dodo Payments. We never see or store your card number; we store your Dodo Payments customer and checkout references, subscription status, and payment and refund records needed to maintain your credit balance.
• 你發送給我們的消息。 你在應用中提交的反饋和發送給我們的郵件。

AI 功能如何使用你的手稿

Anthropic's Claude API processes manuscript passages for AI features. OpenAI's embeddings API processes paragraphs and search queries to support meaning-based manuscript search. Preparing this search index can run in the background after you save or import chapters. Requests use encrypted connections. A full-manuscript analysis may process the entire manuscript within the coverage shown for that scan.

• We do not use your writing to train AI models or opt your manuscript into provider training programs. Anthropic and OpenAI do not train on commercial API content by default.
• We send content needed for the feature and its supporting processing, such as search indexing. Provider retention is separate from model training.
• AI 結果（建議、分析）存儲在您的帳戶中，因此您無需付費重新生成它們。

AI provider retention

We do not promise Zero Data Retention. Anthropic's standard API policy deletes inputs and outputs within 30 days, with exceptions for legal obligations and policy enforcement. Flagged content may remain for up to two years and safety classifications for up to seven years. Feedback deliberately submitted to Anthropic can be retained for five years and may be used for training; we do not submit your manuscript as provider feedback.

OpenAI's standard API abuse-monitoring retention is up to 30 days, with longer retention possible for legal or safety reasons. The embeddings endpoint does not retain application state. Deleting a project in Novilot does not itself erase provider logs or override their retention obligations.

查看 Anthropic's retention policy 以及 OpenAI's API data controls.

Human access

Private means your manuscript is not public or available to other writers unless you share it. It does not mean end-to-end encryption: our servers and processors need readable text to provide the service. People with privileged infrastructure or database access can technically access it. Operational access must be limited to what is necessary for support, debugging, security, abuse investigations, or legal obligations, not browsing your work for personal or commercial use.

Service providers can also permit authorized human access under their agreements, including safety review. We cannot promise that no person will ever review content. Text you choose to include in a support message is available to the people handling it.

分析

We use Google Analytics and PostHog to understand how the product is used so we can improve it. PostHog session replay is configured to mask text and inputs. Sentry helps diagnose errors; request-body collection and default personal-data collection are disabled, and replay text and inputs are masked. Diagnostic metadata may still be processed. We do not intentionally send manuscript text to analytics or error reports. We do not run ads.

Cookie

• nc_工作階段：讓您保持登錄狀態的重要 cookie。沒有它，應用程序就無法運行。
• 來自 Google Analytics 和 PostHog 的分析 cookie，僅用於上述目的。

你的數據存在哪裡，以及誰會處理它

你的數據存儲在託管的 Postgres 數據庫中，並通過 Vercel 提供服務。我們只會為運行 Novilot 的必要目的與服務提供商共享數據：

• Vercel（託管）
• Prisma Postgres（數據庫）
• 人擇（AI 功能）
• OpenAI (manuscript and query embeddings for search)
• Dodo Payments（支付）
• Resend（事務郵件）
• Google Analytics 和 PostHog（分析）
• Sentry (error monitoring)
• Dropbox and GitHub (only when you connect an external backup destination)

We do not sell your data. Manuscript processing is limited to providing the service, your chosen sharing and backup features, and necessary legal or security obligations.

保留與刪除

• Project deletion removes its manuscript, chapter history, Story Bible, summaries, continuity records, embeddings, and saved analyses from the active database. Manuscript-derived voice profiles are invalidated so they cannot keep deleted excerpts.
• Chapter deletion removes that chapter and its version history, and invalidates affected derived analyses and voice profiles. Project-level Story Bible entries, outlines, and notes you have retained remain until you edit them or delete the project.
• To delete your entire account, contact us (see below). Account deletion removes your writing, AI working data, usage history, profile, and linked in-app feedback from the active database. Records required by law, such as payment records held by the payment provider, can have separate retention obligations.
• Our Prisma Postgres plan keeps automated database snapshots for seven days. Deleted data can remain in an earlier snapshot until it expires. This backup window is separate from the AI-provider retention and external copies described on this page.
• Exports, copies shared with others, and backups in your connected Dropbox or GitHub account are not erased by deleting a Novilot project or account. You control those copies separately.
• The browser can hold analysis caches and unsynced edits. Deleting through the app clears the affected local copies on that device. Copies on offline or other devices may remain until cleared there.
• 你可以隨時從應用中導出手稿（DOCX、EPUB、HTML、Markdown 或 PDF）。

你的權利

根據你的居住地（例如 GDPR 或 CCPA 的適用範圍），你可能有權訪問、更正、導出或刪除個人數據，也有權反對某些處理。請聯繫我們；無論你身在何處，我們都會尊重這些請求。

兒童

Novilot 不面向 13 歲以下兒童，我們不會明知故犯地收集他們的數據。

變更

如果我們更改本政策，會更新本頁和頂部日期。如果更改實質性影響手稿的處理方式，我們會在生效前通過應用或郵件通知你。

聯繫我們

郵箱 support@novilot.com，或使用應用程序內的“發送反饋”按鈕。隱私請求由人工回答。