Skip to main content

Sanolith Blog

Writing on HIPAA-aligned LLMs

Engineering posts on PHI redaction, per-tenant fine-tuning, audit trails, and what actually goes into a clinical-grade AI workspace. No fluff. No "AI is transforming healthcare." Just the technical and operational details that matter to teams shipping this work.

8 min readSanolith Engineering

How HIPAA-compliant is ChatGPT Enterprise, really?

OpenAI signs a BAA. That's necessary but not sufficient. Here's what actually happens to PHI in ChatGPT Enterprise, and where the gaps that matter to a privacy officer hide.

HIPAAComplianceChatGPTBuyer guide
7 min readSanolith Engineering

What a fail-closed PHI redactor actually does

Most healthcare AI marketing says 'we have a redactor.' Few say what it catches, what happens when it errors, and why fail-closed is non-negotiable.

HIPAAEngineeringPHI redaction
9 min readSanolith Engineering

Per-tenant fine-tuning without leaking your data

Shared models memorize training data. Per-tenant fine-tuning is the only way to specialize a model for your team without your data leaking to someone else's queries. Here's how it actually works.

Fine-tuningSano adapterMulti-tenant