Responsible AI

Our principles

AI at HealthTasks is built for competency-based education programs that need scale without surrendering academic integrity.

• Augment educators, never replace them

  AI handles first-pass work at scale. Faculty retain judgment, overrides, and final accountability for student outcomes.

• Human oversight by design

  Critical workflows keep educators in the loop. Faculty can review, edit, appeal, and disable AI features when your program needs full manual control.

• Protect educational and clinical data

  AI processing follows the same FERPA- and HIPAA-aligned controls as the rest of the platform, with clear limits on how model providers may use your data.

• Prefer evidence over claims

  We publish validation studies, ROI case studies, and methodology so partners can evaluate how AI performs in real programs. Demos alone are not enough.

Human oversight

Educators remain accountable for assessment quality. AI can draft, score, summarize, or surface patterns. Faculty decide what stands. In skills checkoffs, educators can intervene on appeals or structural divergence rather than re-grading every submission by hand. They still retain full override authority.

Programs can also turn AI features off when institutional policy or accreditation context requires a fully manual path.

Data, privacy, and model providers

We use paid-tier Google Gemini, Vertex AI, OpenAI, and xAI Grok APIs, plus Cloudflare-hosted open-source embedding models. We do not use consumer or unpaid free-tier model services. Under the terms we operate on, providers do not use your data to train their models.

A built-in AI scanner reviews clinical logs for HIPAA Safe Harbor identifiers, including patient names, dates of birth and other patient-specific dates, exact ages 90 or older, medical record and account numbers, encounter or visit identifiers, insurance or member IDs, Social Security numbers, driver's license numbers, phone, fax, email, street address, city and ZIP, room or bed numbers when combined with other identifying detail, device serial numbers, license plates, and other unique IDs. Students cannot save or submit clinical logs until those flags are cleared, preventing PHI and PII from entering our database.

For data routing, retention, identifier screening, and contractual details, see our Privacy Policy and Terms.

Safeguards

Practical controls we build into AI-assisted workflows:

• Educator override and appeal paths on AI-assisted grading
• Optional disablement of AI features for institutions that need it
• Media-quality checks that fail closed when recordings are unsuitable for evaluation
• Institution-controlled environments and role-based access
• Paid-tier providers do not use your prompts or responses to train their models
• Encryption in transit and at rest, with BAAs covering HIPAA-relevant infrastructure
• Built-in AI scanner that flags HIPAA Safe Harbor identifiers in clinical logs; students cannot save or submit until flags are cleared, preventing PHI and PII from entering our database

What we will not do

• Replace faculty judgment as the final authority on student competency
• Train public foundation models on your institutional content without contractual controls
• Present AI output as professional medical, legal, or accreditation advice
• Hide where AI is used in a workflow or how educators can intervene

Evidence

For validation studies, case studies, and related publications, see our Research.

Questions

Contact support@healthtasks.ai or visit our contact page.