BiteJot AI Privacy Policy

Last updated: 12 November 2026. Contact: hardwalker1212@gmail.com. 中文版

Records and identifiers

BiteJot AI supports weight management through food, weight, activity, water and fasting records. These records and coach conversation history are stored on your device. We do not require an account or collect an advertising identifier. The app generates a random installation identifier and an authentication token; our Cloudflare server stores the identifier, a hash of the token, creation and last-use dates, and AI usage records and model-attempt counts for meal suggestions, menu scans and cravings associated with the installation. These are device-linked records, not fully anonymous data.

AI processing and your choice

Only after you agree, photo recognition, AI food parsing and the coach send the content needed for the request through our Cloudflare server to OpenAI GPT-6 Luna through Cloudflare AI. Food input can include a meal photo, typed or dictated food text, and candidate food names. The coach can include your question, recent conversation messages and a profile/recent-record summary: birth year/age, height, supplied sex, weight and trends, goal and target date, intake goal, tracking difficulty, habits, common foods and diet, activity and water summaries. It is not a full raw meal-history export. If you scan a restaurant menu, the menu photos you take are sent the same way, together with the same information used for meal suggestions, and our server does not keep the photos. Meal suggestions on the Eat tab, after you agree to them separately, can include the same summary plus what you logged today, what is left of your goal, patterns worked out on your device such as late dinners, candidate food names, and the tastes you choose to enter: foods you love, foods you avoid or are allergic to, how you usually eat and cuisines. Tastes are stored on your device and are optional. If you use Craving, after you agree to it separately, the food you type, the feelings you pick (such as sweet or crunchy) and whether you are hungry are sent together with the same information used for meal suggestions, to get a few ideas; they are not kept on our server. If you photograph the display of an exercise machine or a workout summary to log an activity, after you agree to it separately, only that photo is sent, to read the activity, duration, calories and distance shown; your records are not sent with it, and our server does not keep the photo. What is read is shown to you before anything is saved. Nothing from Apple Health is sent. Suggestions are stored on your device and can be wrong; they are not medical advice.

Dictation uses Apple Speech recognition, which may use Apple servers depending on device and service availability. We send the resulting text, not the audio, to the AI service. You can decline AI or withdraw consent in Settings → AI & privacy. Manual search, barcode lookup and local matching remain available.

Providers, logs and retention

Our application server does not intentionally persist AI photos, prompts or replies. Its diagnostic events contain status/error categories, counts and timings rather than this content. We have disabled AI Gateway log collection and enabled the gateway's Zero Data Retention setting. That setting applies to eligible Unified Billing calls using Cloudflare-managed credentials; it is not a promise that every provider record or legally required safety record is eliminated. OpenAI states that API content is not used for model training unless the customer opts in; its standard abuse-monitoring retention and safety exceptions can apply according to the service configuration. We do not promise that processing stays in a specific country.

Older AI usage and model-attempt count rows are eligible for opportunistic cleanup after a seven-day cutoff; deletion is not guaranteed on day seven. Anonymous daily model-attempt totals without installation identifiers are retained separately to enforce the shared service budget, including after an installation is erased; they follow the same opportunistic cleanup cutoff. Installation/authentication records have no automatic expiry in the current implementation. Service providers may retain operational and security metadata under their policies. In Settings → Server data, you can copy your random installation identifier or confirm deletion of the current installation’s server authentication, entitlement, AI allowance and model-attempt count records. The request is authenticated by the app; we do not ask you to email a token. Deletion turns off AI and leaves local food, weight and activity records intact. Opening an online feature again may register a new installation. Provider operational records and backups follow their separate retention policies. Contact us for access, correction or other privacy requests; the identifier helps locate records but is not proof of ownership. Deleting the app removes app records from the device but does not automatically delete server records or backups held by you or your platform.

After you delete the current installation’s server data, we still keep the latest status of each Apple subscription involved: Apple’s original subscription identifier, its status (active, expired or refunded), its expiry, the product and trial flag, and a version time used to order updates (Apple’s signing time for states we receive from Apple; for records carried over from before this change, our server’s update time). We use it to check entitlements and to stop an old, refunded transaction from being used again. These records are stored separately from installation records and contain no installation identifier and no food, weight or other records, but the same subscription identifier can be linked to installations that still use that subscription or sync it later. There is currently no automatic deletion of these subscription status records. The flag in your device’s Keychain that records whether this device has used the free coach experience is not removed by this server deletion and is not sent to our server.

Provider information: Cloudflare privacy policy, Cloudflare Unified Billing/ZDR, OpenAI privacy policy, OpenAI API data controls, Apple privacy policy.

Food databases and Apple Health

Online search sends search terms to our server. Barcode lookup sends barcode digits through our server to USDA FoodData Central and/or Open Food Facts. For food logging and matching, nutrition comes from food databases. In version 1.2, restaurant-menu recommendations instead use AI-estimated calorie ranges based on dish names, not database-verified values; ingredients, portions and preparation may differ. Review and adjust estimates before logging. If you allow Apple Health access, the app reads the supported weight, steps, active and resting energy, workouts and sleep data you choose, as supported by your app version, to help calculate energy balance and daily goals. The app does not write to Apple Health. In BiteJot AI 1.1 (20) and later, Apple Health readings and figures derived from them are excluded from the automatically generated summary sent to our server and the AI coach. That summary is calculated from information you enter in the app, independently of Apple Health. Coach conversations from before this change are not included with new questions after you update. Earlier versions, with your AI consent, could include an intake goal calculated with Apple Health activity data in that summary. Content you deliberately type or dictate into an AI request is still processed as described above. We do not use Apple Health data for advertising or marketing.

Subscriptions and purchase information

Version 1.1 adds an optional auto-renewable BiteJot Pro subscription; basic features remain free. Apple processes payment through your Apple Account; we do not receive your card or bank details. To verify access, our server processes signed Apple transaction and notification information, including transaction and original subscription identifiers, product, environment, purchase/expiry and signing times, trial status and refund/revocation status when supplied. Subscription-to-installation associations and the latest subscription status are retained as described above. Restoring a purchase can associate that subscription with another installation on your devices. Deleting app/server data does not cancel an Apple subscription; manage or cancel it in Apple’s subscription settings.

Other uses and your rights

We do not sell your data, use it for advertising or track you across other apps and websites. No advertising or third-party analytics SDK is included. If you voluntarily email support, we receive the information you include and use it to address your request; do not send photos, health records or credentials unless needed and requested. Depending on your location, you may have rights to access, correct, delete or restrict processing, and complain to a data-protection authority. Contact us to exercise them.

The app is intended for adults aged 18 and over. It provides estimates and general wellbeing information, not medical diagnosis or treatment. We will update this page when our practices change.

BiteJot AI 隐私政策

更新日期:2026年11月12日。联系邮箱:hardwalker1212@gmail.com。

记录与设备标识

饮食、体重、运动、饮水、轻断食和助手对话保存在设备上,不要求注册账号,不使用广告标识。App 生成随机安装编号和认证令牌;Cloudflare 服务器保存安装编号、令牌哈希、创建和最近使用时间,以及与安装关联的 AI 使用记录。这些记录关联设备,不是完全匿名数据。

AI 与同意

只有同意后,AI 才把完成请求需要的内容经我们的 Cloudflare 服务器和 Cloudflare AI 发送给 OpenAI GPT-6 Luna。食物输入包括餐食照片、输入或说出的饮食文字、候选食物名称。助手可能包括问题、近期对话消息,以及出生年份/年龄、身高、已填写性别、体重及变化、目标和目标日期、摄入目标、记录困难、习惯、常见食物、饮食/运动/饮水摘要,不是完整逐餐记录。如果你扫描餐厅菜单,你拍的菜单照片也会这样发送,并附带和饮食建议相同的信息,我们的服务器不保存这些照片。“吃什么”页的饮食建议需要你单独同意,可能包括同样的摘要,以及今天记录的内容、还能吃多少、在手机上算出的规律(如晚饭偏晚)、候选食物名称,和你选择填写的口味:爱吃、不吃或过敏的食物、平时怎么吃、口味偏好。口味保存在手机上,可以不填。使用“嘴馋了”需要你单独同意:你写下的想吃的东西、选的感觉(如甜的、脆的)和是否饿,会和饮食建议所用的信息一起发送,用来给出几个选择;我们的服务器不保存这些内容。如果你拍运动器材屏幕或运动总结来记录运动,也需要单独同意:只发送这张照片,用来读出上面显示的运动类型、时长、消耗和距离,不附带你的记录,我们的服务器不保存照片;读出的结果先给你确认,再保存。不会发送 Apple 健康的数据。建议保存在手机上,可能有误,不是医疗建议。

语音由 Apple Speech 转成文字,视设备和服务情况可能使用 Apple 服务器;发送给 AI 的是文字,不是录音。可拒绝或在设置的“AI 与隐私”撤回同意,手动搜索、条码和本机匹配仍可用。

服务商、日志与保留

我们的应用服务器不主动持久化 AI 照片、提示或回答,诊断事件只记录状态、错误分类、计数和耗时。已关闭 AI Gateway 日志采集,开启网关零数据保留设置;后者适用于符合条件、使用 Cloudflare 托管凭证的 Unified Billing 调用,不代表所有运营记录或依法需要的安全记录都不保留。OpenAI 说明 API 内容默认不用于训练,除非客户主动选择;标准滥用监测保留及安全例外按实际服务配置适用。不承诺全部处理在某个国家进行。服务商政策链接见上方英文部分。

超过七天截止日期的 AI 使用和吃什么/菜单/嘴馋了的模型尝试计数记录可被概率清理,不保证第七天删除。不含安装编号的每日模型尝试总数另行保存,用于控制服务的共享预算,不因删除安装记录而重置,适用相同的概率清理截止日期。安装认证记录目前没有自动过期机制。服务商按政策可能保留运营和安全元数据。可在“设置 → 服务器数据”复制随机安装编号,或确认删除当前安装的服务器认证、权益、AI 额度和模型尝试计数记录。App 通过身份认证提交删除请求,不需要通过邮件发送令牌。删除会关闭 AI,保留本机饮食、体重和运动记录;再次主动使用在线功能可能注册新的安装。服务商运营记录和备份按各自保留政策处理。访问、更正及其他隐私请求可联系邮箱;安装编号有助于定位记录,但不能单独证明身份。删除 App 会清除设备内 App 记录,但不自动清除服务器记录、你或平台保留的备份。

删除当前安装的服务器数据后,我们仍会按 Apple 订阅保留最新状态记录,包括 Apple 原始订阅标识、状态(有效、过期或已退款)、到期时间、产品和是否试用期,以及用于排序更新的版本时间(来自 Apple 的状态使用 Apple 签名时间;从这项变更之前的记录转入的,使用我们服务器的更新时间),用于核验权益并防止旧的已退款交易被重复使用。这些记录与安装记录分开保存,不含安装编号或饮食、体重记录,但同一订阅标识可能与仍使用或之后重新同步该订阅的安装关联。目前没有自动删除这些订阅状态记录的机制。手机钥匙串中“这台设备是否用过免费教练”的标记不会因这项服务器删除操作而移除,也不会发送给我们的服务器。

食品数据与健康

在线食物搜索词发送到我们的服务器;条码数字经服务器查询 USDA FoodData Central 和/或 Open Food Facts。食物记录与匹配使用食物数据库的营养信息。1.2 版本的餐厅菜单推荐则由 AI 根据菜名估算热量范围,并非数据库核验值;配料、份量和做法可能不同,记录前应核对并调整。经你授权,App 按当前版本支持的范围读取你选择的 Apple 健康体重、步数、活动消耗、静息消耗、体能训练和睡眠数据,辅助计算能量差和每日目标;App 不会写入 Apple 健康。在 BiteJot AI 1.1 (20) 及以后版本中,发送给我们的服务器和 AI 教练的自动生成摘要不包含 Apple 健康读取值及其派生数字;摘要仅根据你在 App 中填写的内容独立计算,不使用 Apple 健康数据。升级后,这项变更之前的教练对话也不再随新问题发送。更早的版本在你同意 AI 后,可能在摘要中包含由 Apple 健康活动数据参与计算的摄入目标。你主动输入或口述给 AI 的内容仍按上方 AI 处理说明处理。Apple 健康数据不用于广告或营销。

订阅与购买信息

1.1 版本增加可选的 BiteJot Pro 自动续订订阅,基础功能仍免费。Apple 通过你的 Apple 账户处理付款,我们不接收银行卡或银行账户资料。为核验权益,服务器处理 Apple 签名的交易与通知信息,包括交易及原始订阅标识、商品、环境、购买/到期和签名时间,以及 Apple 提供的试用、退款或撤销状态。订阅与安装的关联及最新订阅状态按上文说明保留。在你的其他设备上恢复购买时,该订阅可关联到新的安装。删除 App 或服务器数据不会取消 Apple 订阅,请在 Apple 的订阅设置中管理或取消。

用途与选择

不出售数据,不用于广告,不跨 App 或网站跟踪,没有广告或第三方统计 SDK。主动联系支持时,我们处理邮件中提供的信息以答复请求;不要随意发送照片、健康记录或凭证。根据所在地法律,你可能有访问、更正、删除、限制处理及向监管机关投诉的权利,可通过邮箱提出请求。产品面向18岁及以上成年人,提供估算与一般健康信息,不作医疗诊断或治疗。处理方式变化时会更新本页。