Building Practice around Responsible AI Principles in the Social Sector

On August 19, at 11 ET/4 pm BST, the Ethics and Governance Learning Group at MERL Tech’s CoP is meeting for a session focusing on the state of play for converting responsible AI frameworks to operations and use in social impact sectors. 

AI use is accelerating across philanthropy, international development, and the wider social impact sector, while organizations are still working out what responsible use requires in practice. Teams are making decisions about tools, vendors, data, oversight, and accountability in real time, often without shared guidance or practical resources, sometimes overwhelmed by frameworks and PDFs. This session creates space to compare what organizations are seeing, identify the most urgent gaps, and help shape a useful sector-wide approach that accommodates the nuances of context.

The session will begin with a short, accessible explanation of how AI systems work, where risks and accountability issues can enter, and how the proliferation of responsible AI policies and frameworks can make it difficult to apply principles to practice. We will discuss the current state of play and emerging trends within social impact organizations, exploring the critical questions teams are facing as they weave responsible AI practices into their work. The session will introduce BenevAI, an initiative supporting field-building for the social sector’s AI assurance ecosystem. We will tour the Responsible AI Commons, a publicly available, interactive resource to source and compare responsible AI frameworks, research, and policy documents. Participants will be invited to identify missing perspectives and resources, and contribute feedback through a broader consultation process. The discussion questions below will help shape a collaborative conversation with participants: 

Come ready to discuss: 

• How should organizations decide when and how to use AI and where and how does responsibility sit across funders, implementers, vendors, and affected communities?
• What evidence would your board, funders, or affected community members need to see before trusting an AI-enabled tool or vendor? 
• What is missing from existing frameworks for organisations based in Global Majority contexts?
• When an AI-supported process gets it wrong – how can we know that harm has occurred, what does remedy look like, and who provides it? 
• Are redress pathways currently part of your design and procurement conversations?
• What practical safeguards can organizations apply now, while formal standards are still maturing?

Get to know the speakers: