The honest version of the compliance question: consumer chatbots like ChatGPT and Claude are not HIPAA-covered. Paste patient details into one and you have a compliance problem, not a productivity win. A vendor becomes viable only when it signs a Business Associate Agreement, encrypts data in transit and at rest, and gives you a documented clinician-review step before notes hit the chart. The BAA is the vendor's part; review and consent are yours.

Tools that take this seriously: - Freed, $39 to $119 per clinician per month, the only scribe with fully published pricing. Fast setup, HIPAA BAA included. - Abridge, enterprise pricing on request. Best in KLAS 2026 for ambient AI, built for Epic-native health systems. - Microsoft Dragon Copilot (formerly DAX Copilot), enterprise licensing. Deep Epic and Microsoft ecosystem integration. - DeepScribe, custom pricing. Specialty-tuned models with coding support. - Nabla, free tier up to 30 consults a month. Worth a pilot before any contract. - Heidi Health, free tier available. Popular with solo clinicians. Sources: American Bar Association Start with a BAA in hand, pilot on one clinician, and audit a sample of zero-edit notes against the audio. If a vendor will not sign a BAA, walk away. Teams rolling out clinical AI should run it through a structured PrivateLLM deployment: workshops at $50 plus usage cover the compliance checklist, the rollout, and adoption tracking.