# Tilwe security

We only read your usage. Your prompts stay yours.

We store usage details, never your prompts.

## Only reads usage

OpenAI and Anthropic need an admin key for usage and cost reports. Tilwe only uses it to read those.

## Encrypted at rest

Keys are encrypted with AES-GCM before they touch the database.

## No prompt storage

The gateway records tokens, cost, customer and feature. Never the text.

## Questions

### Do you see my prompts?

No. We store usage numbers only.

### Can I delete my key?

Yes, any time. Or turn on One-time audit and we delete it right after the import.

### Are you SOC 2 certified?

Not yet. We will say so here when we are.

Security questions: team@tilwe.com. Page: https://tilwe.com/security/

## Vulnerability disclosure policy

Tilwe Technologies Private Limited is a small team. We would rather hear about a security problem from you than from an attacker. If you find one, tell us, and we will work with you to fix it.

We do not run a paid bug bounty. We will thank you publicly if you want us to.

### How to report

Email team@tilwe.com with "Security" in the subject. Please include:

- what you found and where (URL, endpoint, or request)
- steps to reproduce it, or a proof of concept
- what an attacker could do with it
- your name or handle, if you want credit

Write in English. If the report contains sensitive detail, say so in a first short email and we will agree a way to send it. Please do not report through social media or GitHub issues.

### Response times

Times a three person team in India (IST, UTC+5:30) can honestly meet. Business days are Monday to Friday, excluding Indian public holidays.

| Step | Our target |
| --- | --- |
| Confirm we received your report | 2 business days |
| First human response with our initial assessment | 5 business days |
| Tell you whether we accept it and how severe it is | 10 business days |
| Fix a critical issue (for example, reading another customer's data or provider keys) | target 7 calendar days |
| Fix a high issue | target 30 calendar days |
| Fix medium and low issues | target 90 calendar days |

We will update you at least every 10 business days until it is fixed. If we will miss a target, we will say so and explain why.

### In scope

- tilwe.com (this site)
- app.tilwe.com (the Tilwe app)
- api.tilwe.com (the Tilwe API)
- the Tilwe AI gateway and any other service run by Tilwe
- how Tilwe stores and uses the provider keys you connect, and Tilwe gateway keys (tlw_...)

We especially want to hear about: one workspace reading or changing another workspace's data; reading or decrypting stored provider keys; bypassing roles; forged webhooks; server side request forgery through webhook channels; authentication bypass.

### Out of scope

- Services run by other companies, even when we use them (Clerk, Razorpay, Google Cloud, Resend, Slack, Cloudflare, OpenAI, Anthropic, Google Gemini). Report those to the vendor.
- Denial of service, load or volume testing, and anything that degrades the service for others.
- Social engineering, phishing or physical attacks on Tilwe staff.
- Spam or abuse of invite and email features at volume.
- Reports from automated scanners without a demonstrated impact.
- Missing best practice headers or settings with no exploit.
- Issues that need a rooted device, a compromised browser, or physical access to a victim's machine.

### Rules for testing

- Only use accounts and workspaces you own. Create a free workspace for testing.
- Never access, change or delete data that is not yours. If you reach someone else's data by accident, stop, do not keep a copy, and tell us.
- Do not use provider keys belonging to anyone else. Use your own keys with low limits.
- Keep request rates reasonable, so you do not affect other customers.
- Give us 90 days from your report, or until we release a fix, whichever is sooner, before you tell anyone else. If we need longer, we will agree a date with you.

### Safe harbour

If you follow this policy in good faith, we will:

- treat your research as authorised by us
- not take legal action against you, or ask law enforcement to act against you, for that research
- not suspend your account for that research
- work with you to understand and fix the issue

If a third party takes legal action against you for research that followed this policy, we will make it known that your work was authorised by us.

This does not cover actions that break the rules above, such as accessing other people's data on purpose, or demanding payment in exchange for not disclosing a finding. We can authorise testing of our own systems only, not our vendors' systems.

Modelled on the disclose.io core terms (https://github.com/disclose/dioterms).

### Recognition

With your permission we will credit your name or handle here once the issue is fixed.