Privacy Policy
Last updated: 2026-08-31 · Operator: ScholarMind AI LLC
中文摘要：使用 AI 工具时，您输入的关键词/文本会发送给模型与数据集服务商（Anthropic、必要时 Google，以及 Hugging Face/UCI/Kaggle）以返回结果；我们会临时记录 IP 用于限流。公开 Guide 使用第一方、无 Cookie 的访问分析：记录 Guide、浏览/有效阅读/按钮类别、来源域名、语言及粗粒度设备类型；IP 只在请求处理中截断并生成按日轮换的密钥哈希，不把原始 IP、完整浏览器标识、查询参数或学生账号写进分析数据库。HealthCom 真实数据图工具会把获准的公开数据源暂存在非公开服务器缓存中（最长约 30 天），只向模型发送题目、数据结构摘要和受控图表选项，不发送原始健康记录；生成图、代码和复现包约 30 分钟后删除。已登录学员打开选题生成按钮时，我们保存账号与题目及汇总次数，供课程管理员联系确认选题；该兴趣记录不保存作者表单或生成文件。科研评估表只收年龄段、不收出生日期，并保存学生主动填写的学术背景、兴趣、学习投入与简答题，以及老师的内部评分和发布反馈，用于项目分层与沟通；13岁以下不接受在线提交，13–17岁必须填写家长或监护人邮箱，18岁及以上可选填。报名课程时我们会收集姓名、邮箱、电话，并由 Stripe 处理付款。若您通过合作代理商或招生内容的专属链接报名，我们会记录该帖子代码、汇总点击与报名归因，并按确认页的明确说明向相应合作方展示本次咨询人的联系方式、Invoice 与付款状态；合作方看不到银行卡信息。我们不使用第三方广告像素，也不要求访问您的社交账号。代理商开户时，我们会保存固定版本合作协议、文本摘要、机构与区域、签署姓名、邮箱、确认时间和安全审计信息，并将其与代理账号绑定。使用联系表时，我们会保存您主动提交的留言及可选姓名、邮箱，以便跟进咨询。我们不出售您的个人信息。您可申请查阅或删除您的数据：mpa@hopeembark.org。
1. Who we are
ScholarMind AI LLC operates ScholarMind AI 科研平台 at academic.boshang.online. Contact: mpa@hopeembark.org.
2. What we collect
| Context | Data | Purpose |
|---|---|---|
| AI tools (topic / analysis / viz / writing) | The text you enter (keywords, topics, notes); your IP address | To generate results; IP is used transiently for rate-limiting/abuse prevention |
| Browser | Local preferences (language) in localStorage; first-party partner-attribution and random visitor cookies after opening a partner link | Remember your language choice and, when applicable, attribute an inquiry to the referring partner. We do not use third-party advertising cookies. |
| Public Guide analytics | Guide id; view, engaged-read, or allowlisted button category; UTC date/time; referrer hostname; page language; coarse device and browser family; a day-scoped HMAC derived from a truncated network prefix | Show approximate, bot-filtered Guide popularity and let administrators compare trends and button usefulness. We do not set an analytics cookie or store raw IP, full User-Agent, URL query/fragment, account identity, or typed page content in the analytics database. |
| HealthCom topic-interest guide (signed-in accounts only) | Account email and display name; canonical topic id/question; first/latest interaction time; aggregate open and successful-generation counts | Let a course administrator see topic interest and contact the student to confirm a research direction. Anonymous visitors are not recorded; author-form fields and generated artifacts are not retained in this record. |
| HealthCom real-data figure generator | Canonical research question; selected public-dataset metadata and a private cached copy of an allowlisted public source; aggregate plotted values; generated figure/code/prompt/manifest; a process-salted account rate-limit key | Compute and validate a reproducible figure. Raw source rows are not sent to the AI model, returned to the browser, or linked to the topic-interest account record. |
| Research-readiness assessment | Age group (not date of birth); student name, grade, school, location, and email; parent/guardian email (required for ages 13–17, optional for adults); self-reported academic background, experience, interests, learning profile, and short answers; teacher scores, internal notes, placement, and published feedback | Apply the age-appropriate submission flow, notify a parent/guardian for ages 13–17, evaluate readiness, recommend an appropriate research level, and provide private feedback |
| Contact / feedback form | Your message; optional name and reply email; source page | Answer your inquiry, route feedback, and keep a service record |
| Course enrollment / payment | Name, email, phone, payment details | Process enrollment and payment; payment is handled by Stripe (we do not store card numbers) |
| Partner-referred enrollment | Referral click and random visitor ID; purchaser/student contact details; fixed enrollment-acknowledgement version, typed signature, timestamp and IP; Stripe Invoice ID, amount and status | Attribute the inquiry, preserve the electronic confirmation, issue and reconcile an Invoice, and let the referring partner follow up. The partner sees contact, Invoice and status only after the purchaser gives the on-page consent; no card number is shared. |
| Recruiting campaign links | Random post code; aggregate click count; the content item associated with an enrollment order | Measure a reviewed recruiting post and attribute enrollment/commission to the correct content and approved partner. We use a first-party redirect, not a third-party advertising pixel, and do not receive social-account credentials. |
| Partner agreement, responsible promoter and payout administration | Invited organization, legal business name, entity type, formation jurisdiction, territory and commercial rate; fixed agreement text, version and SHA-256 digest; authorized signer's name and email; the actual promoter's business contact details, title, relationship to the entity and service location; eligibility/workforce/disclosure attestations and review status; W-9 receipt/verification status; commission ledger, refund adjustments and payout references | Preserve electronic acceptance, verify the business and responsible promoter before activation, attribute and reconcile commission, and administer tax-ready payouts. This portal does not request citizenship or immigration status, identity/work-authorization documents, Form I-9/E-Verify case data, or raw taxpayer-identification numbers. |
合作伙伴登记说明：我们登记实际负责推广的自然人、其业务联系方式、与机构关系、服务地点和机构责任声明，并记录企业、推广人及 W-9 的审核状态。合作伙伴门户不收集公民或移民身份、工作许可/身份证件、Form I-9、E-Verify 案件资料或完整税号；如存在雇佣关系，法定雇主应在独立的人事流程中依法处理。
Most AI tools do not require an account; the resource-intensive HealthCom real-data figure generator requires a signed-in student account for fair-use limits. The HealthCom guide clearly identifies when a topic interaction will be linked to your account for administrator follow-up; anonymous use is not linked to a person. Please do not paste others' personal or confidential data into the tools.
3. Service providers (sub-processors)
To deliver the Service we share the minimum necessary data with:
- Anthropic (Claude API) — receives your tool inputs to generate output. Under Anthropic's commercial API terms, inputs/outputs are not used to train their models.
- Google (Gemini API) — only if enabled as a fallback; receives tool inputs when used.
- Hugging Face, UCI, Kaggle, PhysioNet, Our World in Data, and EPA AirData — receive a canonical public-source request for dataset search or the real-data figure generator. The figure tool uses only server-allowlisted source URLs/files and records source terms and attribution in the download manifest.
- Stripe — processes payments for paid features/enrollment.
- Our local model server (self-hosted Qwen2.5 via Ollama) runs on our own infrastructure and does not send your data to a third party.
4. How we use data
To operate and improve the Service, generate results, prevent abuse, process payments, and respond to your requests. We do not sell your personal information.
5. Retention
- General AI tool inputs/outputs & dataset-search results: cached in server memory only, ~6 hours, then discarded (cleared on restart).
- HealthCom real-data figure jobs: generated PNG/SVG/code/caption/prompt/manifest bundles are stored in a private job directory for about 30 minutes, then deleted. Allowlisted public source snapshots may remain in a private server cache for up to 30 days to avoid repeated large downloads; raw source files are never offered as an artifact. The Agent receives only the canonical topic, schema summary, and approved recipe choices—not raw records. These files are not linked to the signed-in topic-interest record.
- IP addresses: Guide analytics does not persist raw IP. It uses an address transiently to make a truncated, day-scoped HMAC, then discards it. Application rate-limit keys remain in process memory for up to 24 hours. Restricted web-server security/access logs may contain an IP and requested URL for up to 14 days under log rotation; only aggregate Guide counts, with those raw fields discarded, may be backfilled into the analytics dashboard.
- Guide analytics: pseudonymous deduplication keys are kept for no more than 48 hours. Daily Guide, domain and button aggregates may be retained to compare long-term educational interest. We honor browser Global Privacy Control and Do Not Track signals for optional Guide event collection.
- HealthCom topic interests: account-linked aggregates are retained for up to 180 days after the latest interaction, or deleted sooner after a verified deletion request.
- Research-readiness assessments: submissions, teacher reviews, and published feedback are retained for up to 24 months after submission so the student and teaching team can use the recommendation during program planning, or deleted sooner after a verified request unless a longer period is legally required.
- Contact / feedback messages: stored in our first-party feedback queue while needed to respond to and document the request. You may ask us to access or delete a message using the request form below.
- Enrollment/payment records: retained while your relationship with us is active and for the period required for tax/financial recordkeeping, then deleted or anonymized.
- Partner referral records: click attribution and lead records are retained while needed to administer the partner relationship; signed acknowledgements and Invoice/payment records are retained for the applicable contract, dispute, tax and financial-recordkeeping periods. You may request access, correction or deletion where the law permits.
- Recruiting campaign records: draft/approved/published content, approval audit fields, aggregate link clicks and enrollment attribution are retained while needed to operate and evaluate the campaign and administer any associated commission; archived records may be deleted or anonymized when no longer needed.
- Partner onboarding, compliance, commission and agreement records: unused invitation links may expire or be revoked; accepted and countersigned agreement snapshots, responsible-promoter attestations/reviews, commission/refund ledger entries, W-9 verification status and payout references are retained for applicable contract, dispute, tax and legal-recordkeeping periods. Sensitive HR eligibility documents are not kept in this portal. You may request access or correction where the law permits.
6. Your rights
Depending on your location (e.g. China PIPL, EU GDPR, California CCPA), you may have rights to access, correct, delete, or export your personal data, and to withdraw consent. To exercise these, email mpa@hopeembark.org or submit a request below.
7. Children
The Service is used in educational settings that may include minors. We collect the minimum data needed and use an age group rather than a date of birth. The research-readiness form blocks online submissions from students under 13 and directs a parent or guardian to contact us. For ages 13–17, a parent/guardian email and confirmation of the parent's or guardian's knowledge and agreement are required; that address receives a submission notice and published feedback. Adults may submit independently. Schools and program partners remain responsible for any additional consent or notice their policies require. Contact us regarding a minor's data and we will assist promptly.
8. International transfers
Our providers may process data in other countries (e.g. the United States). By using the Service you understand your inputs may be processed by these providers as described above.
9. Changes & contact
We may update this policy; material changes will be posted here. Questions or requests: mpa@hopeembark.org.
This is a good-faith starting template based on the platform's actual data flows. Please have it reviewed by qualified counsel (and confirm the contact email and any PIPL/GDPR/CCPA-specific obligations) before charging users.