Trust Centre
What happens to your information, how our AI is used and supervised, and how our systems are protected. Every statement on this page describes how our systems work today. Where we do not hold a certification, we say so.
People sign off what matters
Every assessment, report, proposal, quote and invoice is approved by our lead human advisors before it reaches you. No timer or fallback approves anything by default.
Never used to train AI
Client work runs only on Anthropic’s Claude, under commercial terms that forbid training models on customer content. Free AI models are used only for public research, never with client or enquirer data.
Nothing leaves unapproved
Our database itself refuses to mark any message, report or invoice as sent unless it was approved first. The rule holds even if our own software has a bug.
No cookies, no trackers
No analytics, no advertising tags, and nothing loaded from other companies, including fonts. Only what you type into our form is collected.
Every legal claim checked three times
Each requirement we state is cited to the official text and checked for facts, against the source itself, and by an adversarial expert review.
Encrypted and locked down
Encrypted connections everywhere, encryption at rest by our database provider, row-level security on every table, and secrets kept out of our code.
Check it yourself
You do not have to take our word for the website’s security. These independent scanners test agentl.in live, right now, and show you the result.
Your information
| What | How we handle it |
|---|---|
| What you send us | Only what you type into our form or email to us, used to answer you and to do the work you ask for. We never sell it or share it for anyone’s marketing. |
| Your documents and reports | Delivered through a private link that is signed so it cannot be guessed, and that stops working after 14 days. Ask any time for a fresh one. |
| AI processing | Client work is processed by Anthropic’s Claude under its commercial terms, which forbid training models on customer content. |
| Who else handles it | A small number of providers for hosting, database, email, secure internal messaging, payments and AI, each chosen for its security practices. Our privacy notice lists them by category; we name them to any client or prospective client who asks. |
| Where it goes | Our providers include companies in the United States. Transfers from the EU and UK rely on adequacy decisions or the European Commission’s standard contractual clauses. |
| How long we keep it | Enquiries that do not lead to work: no longer than 24 months after our last contact. Client files and invoices: as long as Indian tax and accounting law requires, then deleted. |
| Your rights | Ask for a copy, a correction or deletion at hello@agentl.in. If you ask us not to contact you, that is permanent: our systems block every future message to you. |
How we use AI, and how we supervise it
We run our own AI the way we advise clients to run theirs: clear roles, human oversight where it matters, and a record of every check.
- AI agents research, draft and cross-check. They are never presented as people, and every email they prepare says that AI was involved.
- Three independent reviews on anything with a legal or regulatory claim: facts, citations against the official text, and an adversarial expert review. Findings that are not resolved reach our lead human advisors flagged, never silently dropped.
- A manager review before anything is sent. It can send work back or tighten the wording, but it cannot approve, send or change a price.
- Routine messages, such as replies and first emails to businesses, go out after these automated reviews within strict daily limits, and say so. Anything flagged, and anything with a price or a deliverable, waits for a person.
- Prices are calculated by code, not chosen by AI, and an invoice must match the quote you accepted.
- Every review is recorded, so for any piece of work we can show what was checked, by which step, and what changed.
How our systems are protected
- Encrypted in transit on every connection, with HTTPS enforced by the browser (HSTS).
- A strict content security policy on every page: no scripts at all, and nothing loaded from anywhere but our own domain.
- Row-level security on every database table, and encryption at rest by our database provider.
- Only one account can approve work. Every approval is checked against our lead human advisors’ verified account; anyone else is refused.
- Every incoming connection is verified. Payment and messaging notifications are accepted only with a valid signature or secret token.
- Secrets never live in our code. Keys are held in our hosting providers’ encrypted secret stores.
- Every change to our code is type-checked and tested automatically, including tests that confirm the statements on this page still hold.
- Payments are taken by our payment provider. We never see or store card details.
Incidents and reporting a vulnerability
If you find a security problem, write to hello@agentl.in. We will acknowledge it, keep you informed, and not take action against good-faith research that avoids harm to others’ data. Please do not access other people’s information or disrupt the service while testing.
If an incident affects your information, we will tell you without undue delay and notify the authorities within the times the law sets, such as six hours to CERT-In in India and 72 hours under the GDPR.
Certifications
We do not yet hold ISO/IEC 27001, ISO/IEC 42001 or SOC 2 certification, and we will not imply otherwise. When we do, the certificate will be linked here. In the meantime, send us your security questionnaire and we will answer it in writing.