Privacy Policy

Quick Summary

1) Information We Collect

• Account & Contact. Name, email, subscription tier, plan metadata.
• Bot Submissions. Text, links, publicly shared images/screenshots, and context you provide to our bots for analysis.
• Device/Usage. IP address, device/browser type, timestamps, pages, diagnostics, crash reports, performance metrics.
• Transactional. Purchase history, amounts, currency, limited Stripe tokens (no full card numbers stored by us).
• Community/Support. Messages you send to support or feedback you choose to provide.
• Sensitive Personal Information. We don’t ask for SPI; if you submit it, we process it only to provide the requested service and protect users (see Limit Sensitive PI).

Victim-safety first. When users submit links or evidence about scams/abuse, we remove direct identifiers before any awareness use and follow least-privilege access internally.

2) Sources

We collect data directly from you, automatically via the Services, and from service providers acting on our behalf (e.g., payments, cloud hosting, fraud prevention). We may also ingest publicly available signals you point us to (e.g., a public social post URL) to detect impersonation or deception.

3) Cookies & Similar Technologies

We use a small number of essential and functional cookies (e.g., session, security, rate-limit). We do not run behavioral advertising cookies. You can control cookies in your browser. If we add analytics, we will use privacy-preserving settings and update this page.

4) How We Use Data

• Provide, maintain, and improve the Services (including safety detections)
• Authenticate users, prevent fraud, abuse, and security incidents
• Respond to support requests and communicate about the Service
• Generate aggregate, de-identified insights and threat statistics
• Meet legal, regulatory, and compliance obligations

Not legal/medical/financial advice. Outputs are AI-generated safety signals to help you decide; they are not professional advice or certification of fact.

5) Lawful Bases (EEA/UK only)

• Contract — to provide the Services you request
• Legitimate interests — security, fraud prevention, service improvement (balanced with your rights)
• Consent — where we rely on it (you may withdraw at any time)
• Legal obligation — to comply with applicable laws

6) Sharing & Sub-processors

We share personal data only with service providers that process it for us under contracts that require confidentiality and appropriate safeguards. We may also disclose data to comply with law, protect users, or in a merger/acquisition (with notice and choices where required).

See Appendix A for our core sub-processors.

7) International Transfers

We are U.S.-based. When transferring personal data internationally, we rely on appropriate safeguards such as the EU Standard Contractual Clauses with our processors, plus additional measures (encryption in transit, access controls). Where local law requires different mechanisms, we follow those requirements.

8) Retention

We keep personal data only as long as needed to provide the Services, comply with law, resolve disputes, and protect users. You can request deletion at any time (see Make a Request). Aggregated, de-identified metrics may be retained for service integrity.

Your scans. When you run a scan, the raw content you submit — the message text, any image, and the full web address with its path and parameters — is used to produce your answer and is not retained afterward. For accounts with a command center (dashboard), we keep a signal-only record of what the scan found: the verdict, its severity, the website's bare domain (never the full link), which guardian answered, threat-intelligence flags about the site, and the timestamp. This record is never shared, auto-deletes after 90 days, and can be cleared by you at any time. If you switch on "Help others," a separately sanitized threat record (no message, no personal details) may be added to the public feed — see your permissions page for details.

Helping the guardians learn. If you choose to share (either by leaving "Help others" on for a genuine threat, or by tapping "Help improve PRINCE" on a specific result), the raw example — the message, image, or link — may be kept in a de-identified training set for up to 30 days to make the guardians better at real scams, then it auto-deletes. It is tied to a one-way hash, never to your name or account, is used only to improve detection under review, and is never sold or shown publicly. If you don't share, nothing about your scan enters this set.

The confirmed-threat pattern library. A scam that is confirmed real teaches the guardians the most — so when a shared example is confirmed (because it matched a known-scam list, was independently reported by several unrelated households, was verified by our team, or was judged a threat by our deterministic evidence check), a redacted, de-identified version of it may be kept in a durable threat-pattern library with no fixed deletion date, the way antivirus software keeps virus definitions permanently. Before anything is stored, the content is scrubbed: names, emails, phone numbers, web links, dollar amounts, and long number strings are replaced with typed placeholders (such as [NAME] or [LINK]). What remains is the manipulation pattern and the brand or agency the scammer impersonated — never your identity, and never the raw message. Each entry is tied to a one-way hash, never to your name or account; it is used only under human review to improve detection; and it is never sold, never shown publicly, and never used to profile any individual. Only shared examples that are confirmed are added: uncertain or harmless scans never enter it, and if you don't share, nothing of yours does either. We keep these de-identified signatures indefinitely because a scam pattern we can still recognize years from now protects more people over time.

Deleting your account. You can delete your PRINCE account and its data yourself from your account page or the app. This erases your profile, plan access, scan-signal history, Solomon memories, trusted-circle records, survey and feedback responses, and connected devices. Two things are handled separately and by design: (1) Billing records — invoices and payment history held by our payment processor (Stripe) are retained as financial-record and tax law require, and are not deleted by this action; cancel your subscription first at your account so billing stops. (2) Already-shared, de-identified contributions to the public threat feed, the 30-day training set, or the durable confirmed-threat pattern library are not tied to your identity and are covered by the timelines above; because a redacted, confirmed scam pattern is stripped of everything that points back to you, it stays part of the shared shield after your account is gone.

9) Security

• Encryption in transit; limited, logged access in production
• Role-based access control; principle of least privilege
• Backups and disaster recovery for core systems
• Vulnerability and abuse monitoring

No security program is perfect; if we identify a breach affecting you, we will notify you and regulators as required.

10) AI Systems & Model Training

• We use reputable AI providers (currently Mistral AI for analysis and image reading; Together AI for inference and audio transcription; Amazon Web Services for translation, image/video safety analysis, threat-speech scoring, backup AI inference, and voice; and Google, Microsoft Azure, and ElevenLabs for speech synthesis of our answers — see Appendix A) to process the content you submit for safety signals, under data-processing terms; they do not use your content for their own purposes.
• We do not use your personal data to train third-party foundation models. Where a provider allows it, we disable training/retention; otherwise we minimize, pseudonymize, or redact.
• We may train our own lightweight classifiers on de-identified, aggregated patterns to improve threat detection. These training sets are stripped of direct identifiers and are not used to profile individuals for ads or unrelated purposes.
• We keep a durable, de-identified library of confirmed scam patterns (see Retention) to help our own guardians recognize known manipulation over time. Its entries are redacted before storage, tied only to a one-way hash, retained with no fixed deletion date, and used solely under human review to improve our detection — never to profile people and never sold or shared for advertising.

11) Automated Decision-Making & Profiling

Our bots generate risk ratings and flags automatically. These are assistive signals for you; we do not make consequential decisions about individuals (e.g., credit, employment, housing) without human review. You may contest or request human review of any automated output that materially affects you.

12) Your Privacy Rights

EEA/UK users: you can lodge a complaint with your local supervisory authority. We’ll cooperate in good faith with all regulators.

13) Children’s Privacy

The Services are not directed to children under 13, and we do not knowingly collect their data. If you believe a child provided data, contact us for deletion.

14) “Do Not Track”

Browsers may send a DNT signal. Because there’s no common industry response standard, we don’t respond to DNT at this time; we already avoid behavioral ads and minimize tracking.

15) Changes to this Policy

We may update this policy to reflect changes to the Services or law. We post updates here and, if changes are material, we’ll provide additional notice.

16) Contact Us

Email our privacy team at legal@projectprince.ai. If we appoint an EU/UK representative or Data Protection Officer, we will update this section.

Appendix A · Core Sub-processors