AI Safety & Limitations
Responsible-AI controls, prohibited uses, and human accountability.
AI features are designed as assistive tools for authorized users. They do not replace required human review, professional judgment, Provider policy, or applicable legal and regulatory obligations. Public screenshots and examples must use redacted or demonstration information.
Purpose
This page defines the safety boundaries that apply to all True Care System AI capabilities. AI is an assistive layer and must not bypass user permissions, Provider policy, regulatory obligations, or required professional judgment.
Required Controls
Material decisions require review and confirmation by an authorized person.
AI cannot grant permissions or disclose records beyond the user’s role.
AI may not combine or expose data from different Providers.
Only the minimum information necessary for the approved task should be processed.
Relevant AI-assisted actions, approvals, and exceptions should be reviewable.
Users should be able to identify whether an answer is based on system data, documentation, or an AI inference.
Prohibited Uses
AI must not diagnose, prescribe, administer medication, or replace licensed clinical judgment.
AI must not make final breach, employment, tax, billing, or compliance determinations.
Do not submit passwords, OTP codes, private keys, or full financial credentials.
Do not transmit PHI to unapproved external AI systems.
AI must not alter operational records without an authorized, attributable action.
Known Limitations
Recommendations may be incorrect when source records are missing, stale, or inconsistent.
AI can misunderstand context or produce inaccurate statements.
Federal, state, payer, and program requirements may change and must be independently verified.
The authorized user remains responsible for the final action.
Security and HIPAA Expectations
Users may access only the records and functions permitted by their assigned role and Provider scope.
Only information reasonably required for the approved operational purpose should be processed.
Authorized personnel remain responsible for validating recommendations and approving final actions.
Material AI-assisted activity should remain attributable, reviewable, and connected to the source workflow.