Trust
How Darkmeter is designed to handle data.
These are the principles Darkmeter is being designed around. They describe intent for a product in development, not certifications or audits.
What goes where
Only what it needs to recognise a person.
Darkmeter is a gate, not a proxy. It is told who is asking and how much was used, and it never sits in the path of the request.
Darkmeter receives
- The email address you send with a login
- A device token from the browser SDK
- The IP address you send
- Credit counts and timestamps
- Your own decisions: block, release, mark as real or abusive
Designed to stay with you
- Prompts
- Replies
- Files and images people generate
- Your AI provider keys and traffic
Requests stay with you
Your product calls Darkmeter before and after an AI request. The request itself goes straight to your provider, so prompts and replies do not reach us.
Metadata only
Darkmeter needs to know who, when and how many credits. It does not need the content of what people ask for.
Isolated per customer
Each customer's data is kept separate. Sharing signals across customers would be a later feature, and only with agreement.
What is still open
Not finished, and said plainly.
Device data and stored emails are likely personal data in many places. We would rather list what is unsettled than imply it is done.
- Retention and deletionBeing defined
- Where data is storedBeing defined
- Consent wordingBeing defined with legal review
- Certifications and auditsNone yet
Build Darkmeter with us.
We are working with a small group of AI teams. Tell us about your free tier and what it costs you.