Using Purple with University Data by Classification


Overview

This guide helps University of Washington employees, faculty, and students understand how to properly handle University data when using Purple, the University's generative AI platform. This guide explains how to use Purple responsibly while safeguarding those records.

When to Treat Purple Outputs as Protected Information

Protect all sensitive information and what Purple creates from it. When you use Purple with protected or sensitive information, the content Purple generates may also contain protected or sensitive information. This includes summaries, analyses, reports, and other content created from FERPA-protected education records, health information, research data, or other confidential University information.

This applies to:

What this means: AI-generated summaries, reports, and other outputs can be just as sensitive as the original information. Treat Purple's outputs with the same protections as the source of information, including the same requirements for access, sharing, storage, and retention.

What Data Can Be Used with Purple

Data classification levels and examples of data allowed with Purple
Class Data type Examples Approved for Purple? Notes / where to get guidance
Level 1: All Audiences Public information Published web content, public policies, and course catalogs Yes Verify outputs before use.
Level 2: Limited Audience Internal, non-sensitive institutional data Internal procedures, meeting notes without personal or sensitive data Yes  
Level 3: Restricted Audience FERPA-protected education records Grades, academic standing, enrollment, and discipline records Yes, with conditions Only access and share student records when you actually need them for your job.
Level 3: Restricted Audience Other sensitive institutional data Confidential data not listed in Level 4 No If unsure, ask your supervisor or the Privacy Office.
Level 4: Special Handling Requirements HIPAA-protected health information Patient records, diagnoses, and billing tied to care No See the UW Medicine Gen AI Policy. 
Level 4: Special Handling Requirements UW Medicine data Clinical, financial, research, and other protected information No See the UW Medicine Gen AI Policy. 
Level 4: Special Handling Requirements Confidential research data Unpublished human-subject data, sponsor-restricted data No  
Level 4: Special Handling Requirements Controlled Unclassified Information (CUI) Federally designated controlled information No  
Level 4: Special Handling Requirements Sensitive personally identifiable information (PII) SSN, financial account numbers, government IDs No  

Table citation: Draft Updates: UW Data Classification Scheme - Information Technology

Key Principles

What is FERPA?

FERPA (Family Educational Rights and Privacy Act) is a federal law that protects the privacy of student education records. It covers information like:

Under FERPA, the University must keep student records private and only share them with people who have a legitimate educational reason to access them.

Important distinction: This guideline covers FERPA-protected Education Records. Treatment records, including records maintained by counseling centers, mental health services, or health care providers for treatment purposes, are subject to different legal and privacy requirements and are not addressed in this document. If you work with treatment records, consult your department or the UW Privacy Office using Purple with that information.

Purple Can Be Used with FERPA-Protected Education Records

FERPA-protected education records may be used with Purple. As with any FERPA-aligned technology, you are responsible for handling both the data you provide and the output you receive appropriately. Student information and Purple-generated content must be shared, stored, and accessed only for legitimate educational purposes and in accordance with applicable University policies and procedures.

Do:

Don't:

Requirements for Using Student Information in Purple

When you use Purple with student education records, you must follow these five key requirements:

1. Use Student Information Only for Legitimate University Purposes

Student information is only for direct work supporting their education at UW. This includes:

What this means: Only access and share student records when you actually need them for your job. Don't use them out of curiosity or convenience.

2. Do Not Share Education Records Inappropriately

Student records are confidential. Do not share them with anyone who doesn't have a legitimate need to know, including:

What this means: Keep student information within the people and departments that need it for specific University purposes.

3. Verify Information Before Acting on It

You must check all AI-generated content before using it. Review what Purple produces for accuracy, context, and completeness before using it for decisions or sharing it with students.

What this means: Use your professional judgment. If the output seems off or incomplete, verify it independently before acting on it or sharing it.

4. Do Not Delegate Important Decisions to AI

Use Purple to help you work more efficiently, but never rely on it as the only basis for decisions that affect students, such as:

What this means: Always add your own professional expertise and judgment. Make final decisions yourself, using Purple as a tool to assist you, not a replacement for your decision-making.

5. Maintain Professional Judgment and Accountability

You are responsible for ensuring that student information is handled carefully and that University privacy policies and FERPA requirements are followed. This includes:

What this means: Follow these guidelines because they protect students and the University. When in doubt, reach out to the Privacy Office or your supervisor.

Examples of What Not to Do

Example 1: Exposing Student Records to Unauthorized People

Situation: A staff member works in the Registrar's office and has access to student records, including academic standing, grades, and discipline history. They ask a colleague in another department to review a student's records in Purple to help identify trends.

Why this is a problem: The colleague doesn't have a legitimate educational need to access these records. Sharing them violates FERPA and University policy.

What to do instead: Only share student records with people in your department who need them for their specific job duties.

Example 2: Using AI-Generated Summaries of Sensitive Information Without Review

Situation: A department head asks Purple to summarize and analyze student records to identify at-risk students. They take Purple's list of recommendations and send it to other staff members without first reviewing it.

Why this is a problem: The AI might have missed important context or made errors. More importantly, sharing outputs based on student records requires the same care as sharing the original records.

What to do instead: Always review AI-generated outputs yourself. Verify the information and think about whether it's complete and accurate before sharing it with anyone else.

System Information and Limitations

What Data Cannot Be Used with Purple

Purple is NOT approved for use with:

UW Medicine Data (see UW Medicine Gen AI Policy for definitions of clinical data, financial data, research data, and other protected information)

Other sensitive institutional data

Important: Check with your supervisor or the Privacy Office if you're unsure whether information is safe to use with Purple.

System Status and Availability

Purple is currently a prototype system and is not approved for permanent production use. This means:

Important Terms and Conditions Agreement

Important: By using Purple and other UW software applications, you formally agree to follow the terms and conditions in the Classes and Agreement for UW Data and Information Systems. These terms cover your responsibilities and the University's responsibilities when handling data.

Current Guidelines and Future Updates

The UW AI Task Force is developing comprehensive guidance for the responsible and ethical use of artificial intelligence (AI) at the University of Washington. Until that guidance is finalized, please follow the interim guidance below.

Interim Guidelines

UW is developing University-wide best practices for the use of artificial intelligence (AI). Until that guidance is finalized, follow these interim principles:

Purple System Updates and Support

Training and Support

AI Guidelines at UW and UW Medicine

All UW users should review:

These documents provide additional context and guidance for the responsible use of generative AI.

Additional Resources

These documents provide additional context and guidance for the responsible use of generative AI.

Questions?

If you have questions about:

Please reach out to the UW Privacy Office, your department supervisor, or UWIT Support.