For companies · Privacy Filter
Use Claude, GPT and Gemini without sending names, addresses and account numbers.
Your data protection team says no because prompts contain personal data? We put a Privacy Filter in front of the major providers' APIs. It replaces this data with placeholders before a request goes out and puts it back into the response.
The provider only sees placeholders.
Click through a request, from prompt to response.
- On your sideYour applicationChat, support tool, script – whatever you already use.
- Run by SeibertPrivacy FilterSwaps detected data for placeholders, there and back.
- With your API keyAnthropic · OpenAI · GoogleThe model of your choice, under your own contract.
The filter marks what it detects.
Your contract, your key. We only run the filter.
You keep the relationship with the provider. We take care of what happens in between.
You
Sign the contract with Anthropic, OpenAI or Google and bring your own API key. Model choice and billing with the provider stay with you.
We
Seibert Group runs the filter model between your applications and the provider. It's built for high throughput and works in both directions.
The provider
Receives your request without the detected data and responds with the same placeholders. It doesn't see the originals.
What the filter detects
Out of the box, the model detects eight types of personal and confidential data. We agree with you on terms that are sensitive only in your context.
To be honest
No filter catches everything. The Privacy Filter minimizes what goes out – it doesn't replace a data protection review. That's why we test it with examples from your day-to-day work before launch.
- Names
- Addresses
- Email addresses
- Phone numbers
- Account and card numbers
- Private URLs
- Dates
- Passwords and keys
- + your own terms, e.g. project names
Frequently asked questions about the Privacy Filter
Is this anonymization within the meaning of the GDPR?
It depends on your use case – that's best clarified with your data protection team. We provide the groundwork.
Does it work with German texts?
The model is trained mainly on English. We use your sample data to check in advance how well it handles your German texts.
Who has the contract with the provider?
You do. You get the API access from Anthropic, OpenAI or Google yourself. Your agreement with us covers only running the filter.
What does it cost?
That depends on the volume and on your own terms. Get in touch and we'll send you a quote.
Tell us what you want to filter.
A few details are enough. We'll get back to you by email with next steps, with no obligation and no login.
Prefer to reach out directly? Write to gpt@seibert.group.