protected class
This glossary entry explains protected class for AI governance and model risk programs. The sections below summarize what the term means in plain language, why chief AI officers and cross-functional committees track it, where teams often get confused, and—when you are signed in—how it shows up across major industries and in expectations tied to the EU AI Act and NIST AI RMF. Use related links at the end of the page to explore neighboring concepts without losing context.
What It Means
A protected class is a group of people with shared characteristics (like race, gender, age, or disability status) who are legally shielded from discrimination in employment, housing, lending, and other areas. When building AI systems, companies cannot use these characteristics to make automated decisions that could disadvantage these groups. This applies even if the AI doesn't directly use protected class data but relies on proxy variables that correlate with them.
Why Chief AI Officers Care
AI systems that discriminate against protected classes expose companies to lawsuits, regulatory fines, and reputational damage under civil rights laws. CAIOs must ensure their models don't inadvertently learn biased patterns from training data or use seemingly neutral features that actually serve as proxies for protected characteristics. This requires ongoing monitoring and testing of AI outputs across different demographic groups to prove fair treatment.
Real-World Example
A bank's AI lending system was trained on historical loan data and began approving fewer mortgages in zip codes with predominantly minority populations, even though race wasn't a direct input variable. The system had learned that certain geographic and economic factors correlated with race, creating illegal discrimination that resulted in a $25 million settlement and required the bank to redesign its entire AI approval process.
Common Confusion
Many people think they're safe from discrimination issues if they simply exclude protected class data from their AI training sets, but this ignores proxy discrimination where other variables like zip codes, school names, or shopping patterns can serve as stand-ins for protected characteristics.
Industry-Specific Applications
See how this term applies to healthcare, finance, manufacturing, government, tech, and insurance.
Healthcare: In healthcare AI, protected classes include patients' race, gender, age, disability status, and other characteristics th...
Finance: In finance, protected class considerations are critical when deploying AI for credit decisions, insurance underwriting, ...
Premium content locked
Includes:
- 6 industry-specific applications
- Relevant regulations by sector
- Real compliance scenarios
- Implementation guidance
Technical Definitions
NISTNational Institute of Standards and Technology
"[a feature] that may not be used as the basis for decisions [and] could be chosen because of legal mandates or because of organizational values. Some common protected [classes] include race, religion, national origin, gender, marital status, age, and socioeconomic status."Source: MIT_Protected_Attributes
"A group of people with a common characteristic who are legally protected from [...] discrimination on the basis of that characteristic. Protected classes are created by both federal and state law."Source: Practical_Law_protected_class
Explore more glossary terms
Discuss This Term with Your AI Assistant
Ask how "protected class" applies to your specific use case and regulatory context.
Start Free Trial