Show HN: Brianni – AI chat on GPT/Claude/Gemini that we can't read (provable) (brianni.co)

Brianni-AI has launched a groundbreaking chat application that integrates major AI models—GPT, Claude, and Gemini—while ensuring user privacy to an unprecedented level. The app's architecture is designed to enforce "operator blindness," meaning even the creators cannot access users' chat conversations. This claim is backed by a provable system: users' conversations are encrypted with keys that the operators do not possess, and the plaintext only exists within a secure AWS Nitro Enclave, which users can independently verify for authenticity. This approach significantly raises the bar for privacy claims in AI, providing a measurable way for users to confirm that their data is indeed protected. Moreover, Brianni-AI incentivizes user participation in validating their system by rewarding the first five individuals who successfully reproduce the attestation measurements. Despite the safeguards against data reading, the architecture acknowledges the reality of metadata collection, ensuring transparency about what can and cannot be seen by operators. This innovation not only addresses common privacy concerns in AI applications but also challenges the industry by setting a new standard for trust through verifiable security measures, potentially reshaping user expectations of data privacy in the AI/ML community.