Work
Seth - Privacy-Preserving Therapy LLM
Overview
Seth is a privacy-preserving therapy LLM currently in beta. Unlike retrieval-only systems, Seth actually learns from conversations while keeping user data private through blockchain-based anonymization. Mental health support that improves through use - without surveillance.
The Problem
AI wellness and therapy tools face a fundamental tension: models improve by learning from conversations, but therapy conversations are deeply private. Users share their most vulnerable moments - trauma, fears, relationship struggles, mental health challenges.
Current options force users to choose:
- Use public AI - Get better responses, but your intimate conversations train someone else's model
- Stay private - Keep your data safe, but miss out on AI that could actually help
Neither option is acceptable for mental health support.
The Innovation
Seth uses Ethereum network protocols to mask user PII (personally identifiable information) while still allowing the model to learn from conversations. The blockchain layer creates cryptographic separation between identity and content.
The result: A system that gets smarter from every conversation without ever knowing who had them.
How It Works
- User has conversation - Normal therapeutic dialogue with Seth
- Anonymization layer - Ethereum-based protocols strip identifying information
- Learning contribution - Anonymized patterns contribute to model improvement
- No PII storage - Identity information never persists, never trains, never leaks
Users get personalization and improvement without the surveillance bargain that defines most AI services.
Why This Matters
Mental health support shouldn't require surveillance. The current AI landscape treats user data as the product - free services in exchange for training data. For therapy, this model is broken:
- Therapy requires trust
- Trust requires privacy
- Privacy and "learning from your data" seem incompatible
Seth proves they're not. Privacy-preserving AI is possible. You don't have to trade your most vulnerable moments for better service.
Technical Architecture
| Component | Technology | |-----------|------------| | LLM | Custom architecture | | Privacy Layer | Ethereum network protocols | | Backend | Python |
The Ethereum integration isn't about cryptocurrency - it's about using blockchain's cryptographic primitives for identity masking. The same technology that enables anonymous transactions enables anonymous learning contributions.
Broader Implications
Seth demonstrates a model for ethical AI development:
- Users contribute to improvement without becoming the product
- Privacy is architectural not just policy
- Trust is verifiable through cryptographic guarantees
- Mental health AI can exist without exploitation
For clients in healthcare, wellness, or any privacy-sensitive domain, Seth shows Blacksky's capability to build AI that respects user boundaries while still delivering value.
Project Type: Product / Beta Status: Beta - Blacksky is raising capital for this project Domain: Mental health / Wellness AI