Principles
Ethical AI Policy
Effective 13 October 2026 · Version 1.0 · Revision history
The rules this tool is built to, and the places where you should not trust it.
1.Lawful open-source intelligence only
A private eye's lawful craft is public record work: registries, court and corporate filings, company filings, domain and web history, public professional profiles. The tool is scoped to that kind of material and to nothing else. Scope choices narrow the report; none of them widen the methods.
2.What the tool will not do
- No hacking, credential testing, or access to any account, inbox or private message.
- No pretexting, impersonation or social engineering to draw information out of anyone.
- No covert surveillance, geolocation tracking or real-time monitoring of a person.
- No pulling from protected or restricted record sets, including credit, medical, criminal-history or protected-government databases.
- No inference about protected characteristics such as race, religion, sexual orientation, health, disability or union membership.
- No output framed as a verdict of guilt, a criminal accusation or a claim of proof.
3.Every finding is a lead, with its receipt
Each finding carries a confidence rating and the sources it was drawn from. That is deliberate: it lets you check the claim rather than inherit it. A finding you cannot open and confirm is a finding you should not use. Red flags are prompts for further enquiry, not conclusions about a person's character.
4.Identity matching is the biggest failure mode
Names collide. The AI can attach a public record belonging to a different person with a similar name, especially where you supplied only a name and a city. Add the identifiers that make a subject unique, such as date of birth range, registry ID, email or an exact domain, and expect to disprove matches as often as confirm them.
5.What this app cannot see
The AI writes from what it already knows. It does not query registries, court roll databases or search engines live, and its knowledge has a cut-off. So an empty section does not mean nothing exists, and a stale filing may have been resolved since. Treat the dossier as a starting index, then run the checks against the real source.
6.Purpose limits and human decision-making
Research a subject because you have a legitimate reason to, not because you can. Output must not drive an automated decision about a person, and any significant decision affecting someone's livelihood, housing, safety or reputation needs a named human who reviewed the underlying sources and can explain the outcome.
7.Transparency and a paper trail
Reports are timestamped, marked as AI-assisted, and keep a version history. Re-running an investigation saves the earlier findings rather than destroying them, so a record of what was known and when survives. Labels and notes on versions are yours, and are recorded as your commentary.
8.Data minimisation and retention
Enter only the identifiers you genuinely need to identify the subject. Cases live in your own browser, are not attached to an account, and are not distributed anywhere by the app. Delete what you no longer need; a dossier about a person is personal data wherever it sits.
9.Safeguarding and emergencies
Nothing here substitutes for a safeguarding response. In an emergency, or where someone is at risk, contact your local emergency services or a relevant safeguarding authority.
The binding terms sit in the Terms of Use. Where the two differ on a point of law, the Terms of Use govern.
Revision history
| Version | Date | Changes |
|---|---|---|
| 1.0 | 13 October 2026 | First published. |