What We Know about Responsible AI Practices in Industry: A Half Decade of Empirical Research

Awareness is growing but implementation lags behind: While practitioners increasingly recognize responsible AI as important, the paper's synthesis of 161 empirical studies reveals that awareness has not translated into consistent practice across the industry. This gap between knowing about RAI and actually doing it remains a central challenge.

Organizational support and resources are unevenly distributed: Practitioners face substantial barriers including limited training opportunities and inconsistent organizational backing for RAI work. This means that responsible AI practices may depend heavily on which company you work for, rather than being standardized across the industry.

Generic tools and guidelines aren't enough: Although toolkits and guidelines have become more widely adopted, the research shows these interventions often don't match how practitioners actually work day-to-day. More tailored, practical solutions are needed to help teams integrate RAI into their regular workflows.

RAI work is becoming more professionalized: The paper documents meaningful progress, including increased practitioner awareness and more formalized RAI activities within organizations. This suggests the field is maturing, though challenges remain in making these practices effective and sustainable.

Industry realities should inform policy: The research emphasizes that policymakers need to understand the actual constraints and contexts practitioners face when designing governance efforts, rather than assuming ideal conditions for implementing responsible AI practices.