@ItsTwoBags@svpino @florian_georg Context from your uploaded documents is still sent to ChatGPT with your request/prompt, but the sensitive data has been redacted before sharing it with ChatGPT. In most cases the redacted PII, PCI, or PHI isn't essential for ChatGPT to product a useful answer.
@ThanhBQuach@svpino Regarding getting the best answer, in most cases the redacted PII, PCI, or PHI isn't essential for producing a useful answer. CapeChat still provides context to ChatGPT from your private documents, but the sensitive data has been redacted from the context.
@ThanhBQuach@svpino Regarding privacy, CapeChat uses secure enclaves for processing requests, so nobody at Cape (or anywhere else) can see the un-redacted data that you submit through CapeChat.
@fdondi1@svpino CapeChat uses secure enclaves for processing requests, so nobody at Cape (or anywhere else) can see the un-redacted data that you submit through CapeChat.
At @capeprivacy, we help companies protect their sensitive data so it can be used in AI pipelines. Hereโs a short demo of how you can use Cape with a @huggingface model to secure both the model and the input data. And if youโre at #PyConUS2023 this week, come see us at booth 411!
This is a simple example that shows how a developer can use Cape Privacy to secure PII or PCI data by encrypting in the browser, and then only processing in a secure enclave.