José Luis Castro is WHO’s Director-General Special Envoy for Chronic Respiratory Diseases and founder and former CEO of Vital Strategies.
Imagine that a domestic violence hotline wants to explore using artificial intelligence (AI) to streamline caller triage. That sounds like a good idea on the surface. But there’s a critical question the team should ask: “Would speed compromise safety?” For instance, if trained on past call data, it could end up deprioritizing non-English speakers because those calls can take longer to resolve. Or it might classify repeat callers as lower priority, despite those callers being at escalating risk levels.
When it comes to using AI in the nonprofit world, what can look like efficiency can lead to systemic biases. There’s a new ethical frontier for AI in civil society. In a sector grounded in equity, trust and human dignity, the question is no longer whether nonprofits should adopt AI, but how to do so with integrity.
Why AI Is Not Neutral
AI isn’t a blank slate. Its algorithms are trained on historical data that, if unchecked, can reinforce the very inequities that many nonprofits set out to dismantle.
Wanting to achieve greater efficiency is often the reason for AI adoption. In the nonprofit world, enhanced efficiency can come in various forms, such as automating intake and forecasting donor behavior. But efficiency alone can be a dangerous metric. Nonprofits are accountable to impact, inclusion and ethical stewardship, which AI can thwart if used improperly.
When AI shapes who receives help first, who is excluded and who designs the rules, these are not backend engineering choices. They are leadership decisions with ethical consequences.
Three Foundational Guardrails
To ensure that AI use aligns with mission and trust, nonprofit leaders must set ethical parameters before adoption, not after harm.
Specifically, they should prioritize three foundational guardrails:
• Transparency: Stakeholders deserve to know when and how AI is used. This means that nonprofit leaders should provide plain-language disclosures, publicly document tools and be clear about data collection practices. Transparency builds trust and enables informed participation.
• Bias Mitigation: Nonprofit leaders should audit datasets and model outputs regularly. It’s vital to test systems across demographics and scenarios and pause deployment if disparities appear. As the hotline example shows, bias is often subtle and deeply embedded.
• Accountability: AI tools need human oversight. Nonprofit leaders should assign responsibility for reviewing outcomes, responding to concerns and intervening when needed. AI should assist—never replace ethical judgment.
Governance As A Leadership Responsibility
AI governance is not just a technical concern. It’s a strategic and fiduciary one. Boards must treat AI adoption with the same rigor as financial oversight or legal compliance.
Specifically, boards should:
• Ensure AI decisions align with the mission and values of their organizations.
• Request documentation on intended use cases, risks and risk mitigation plans.
• Approve ethical use policies before pilots begin.
• Monitor ongoing implementation through regular reporting.
An AI ethics statement is an important part of a strong approach. It can define responsible use, technologies to avoid and how affected communities will be involved. Some organizations might also opt to form cross-functional ethics committees that include staff, leadership, technologists and community representatives. Together, they can review AI tools before and after deployment.
Embedding Lived Experiences Into Design
Data is not objective. If training data reflects only institutional perspectives—or excludes those closest to the problem—it will replicate blind spots.
Community engagement must go beyond consent forms. Nonprofit leaders and their teams should co-design tools with those affected by them by:
• Involving service users in shaping AI features and priorities.
• Reviewing outputs with community representatives before scaling.
• Creating feedback channels to flag harm or bias.
From Policy To Practice: Examples Of Ethical AI Adoption
When it comes to ethical AI adoption in the nonprofit sector, three examples come to mind.
First, there’s Benetech, the U.S.-based education nonprofit. Benetech is creating an “AI-powered platform that transforms teaching materials, especially STEM, into interactive, accessible content. Students with dyslexia or visual impairments will be able to read, listen, and ask questions about challenging concepts like equations and images.”
There’s also DataKind, a global nonprofit that provides data science and AI expertise to other nonprofits. The organization emphasizes its ethical approach to data and AI on its website.
Finally, there’s Thorn. According to a 2024 article in the Los Angeles Business Journal, Thorn “launched an AI-based platform called Safer Predict. The platform aims to help content-hosting platforms (like social media or video streaming websites) detect if their content contains child sexual abuse or may lead to exploitation and grooming.”
AI With Integrity Checklist
Before moving AI implementation from the pilot to scale stages, nonprofit leaders should ask themselves several questions:
• Have we defined the problem clearly—and confirmed AI is the right tool?
• Is the dataset inclusive, representative and reviewed for bias?
• Have affected communities shaped the tool’s design and goals?
• Is the model explainable to nontechnical audiences?
• Is there clear human oversight?
• Do we have plain-language communication for users and partners?
• Are bias audits and reviews scheduled post-launch?
• Does the tool reinforce our equity goals?
Nonprofit leaders should consider AI implementation as an ongoing process that evolves with community input, regulations and organizational learning.
Leading With Purpose In The Age Of AI
AI adoption is a test of values. For nonprofits, the decision carries implications far beyond operational efficiency. It touches the core of trust, inclusion and ethical service—the reasons the sector exists in the first place.
Nonprofits are accountable not just to stakeholders or donors, but to communities whose dignity and rights must be upheld. When AI systems make decisions that affect housing, health, safety or legal protection, etc., those choices reflect ethical commitments—or the lack of them.
AI governance must be proactive and principled. Executives and boards need to treat AI oversight as strategic risk management and community trust stewardship. And most importantly, those most affected must not only be protected from harm—they must be empowered to help shape the technology itself.
In an age of intelligent systems, it’s time for the nonprofit sector to show what intelligent ethics look like: anchored not in speed or scale, but in care, accountability and shared humanity.
Forbes Nonprofit Council is an invitation-only organization for chief executives in successful nonprofit organizations. Do I qualify?
