AI Safety

The lower-right AI chat is limited to questions about Vytherix, this website, our services, workflow fit, privacy, terms, and AI safety. It is not a clinical, legal, compliance, emergency, or customer-support system for existing client operations.

Do not enter protected health information, patient records, medical facts about an individual, credentials, payment data, API keys, confidential client material, or other sensitive information into the public chat.

• Local screening for off-topic, sensitive, and emergency-style prompts.
• Server-side model calls so Gemini API keys are not exposed in the browser.
• System instructions that restrict answers to Vytherix and website topics.
• Refusals for medical, clinical, legal, financial, emergency, or PHI-related requests.
• Short conversation history and message-length limits.

AI can support workflow discovery, drafting, triage, and repetitive administrative steps, but care organizations should keep humans responsible for care decisions, compliance decisions, staffing approvals, vendor approvals, and final operational judgment.

• Start with a workflow audit before selecting tools.
• Use approved systems and documented access controls.
• Minimize data sent to vendors and avoid unnecessary PHI exposure.
• Use BAAs and vendor review when health data may be involved.
• Pilot narrowly, measure outcomes, train staff, and document handoff procedures.

The website and chat cannot help with emergencies, urgent patient care, clinical decisions, or crisis situations. Use emergency services, licensed clinicians, internal escalation policies, or the appropriate local authority.