AI In Philanthropy Series #3: Adopting with Caution and Governance

Adopting AI with Caution
While AI enhances the work of funders and nonprofits, it also has its share of risks and drawbacks – hence, the hesitation of funders and nonprofits to fully embrace it. As CEP states in its 2025 report, “Most foundations and nonprofits use AI in their work but share a common set of concerns about the technology related to security, accuracy, staff expertise, and bias.”

The very speed, scale and complexity that make AI such a powerful tool can also render it harmful when misused due to lack of training and usage guidelines, or to malicious or misaligned intent.

Randima Fernando, co-founder of the Center for Humane Technology (CHT), notes that even well-meaning people get trapped in misaligned incentives. For example, social media heavily relies on AI for various functionalities like content recommendation, personalized advertising, content moderation and content creation. AI algorithms analyze user data, identify patterns and make predictions to optimize the user experience and maximize engagement. As companies and organizations compete for market share through influencers and online advertising, they can engage in harmful behavior, not because they want to – but because they’ll lose if they don’t. He refers to this phenomenon as the “race to the bottom.” The resulting harm, he says, can include:

• Information overload
• Doomscrolling (the compulsive act of endlessly consuming negative and distressing online information)
• Influencer culture
• Shortened attention spans
• Deepfakes (videos of people in which their faces or bodies have been digitally altered so that they appear to be someone else)
• Polarization
• Cult factories
• Online harassment
• Breakdown of democracy

Additional drawbacks of AI are that errors and misinformation can unintentionally occur in its outputs, and over-reliance on it for routine tasks raises concern that it negatively impacts critical thinking skills.

Within philanthropy, key challenges and concerns surrounding AI include:

• Bias. While AI systems can help surface patterns that suggest inequities, they cannot independently eliminate bias. They are only as reliable as the data on which they are trained and the oversight mechanisms governing their use. Consider, for example, how this impacts social justice efforts. Inaccurate AI can perpetuate and even amplify existing stereotypes, discrimination and inequity. An AI tool trained on Western, English-language content might unintentionally misrepresent or exclude marginalized communities, leading to ineffective messaging and unfair funding decisions.
• Integrity and authenticity. As AI-generated content can be indistinguishable from human-created content, funders are increasingly skeptical when reviewing grant proposals. From its 2024 survey of foundation giving trends, Candid learned that 57% of grantmakers don’t know whether they’ve received applications created with AI. When asked if they would accept AI-generated proposals, 23% of grantmakers said no, 67% were undecided and just 10% said yes. Similarly, in 2023 we asked a small sample of grantmakers if they were using GAI in their philanthropic planning and 48% said definitely not, 26% had no immediate plans to do so, 13% said not yet but they plan to, and 13% said yes.
• Data privacy. As many charitable giving entities handle sensitive and confidential material such as donor data and beneficiary information, using AI to analyze it raises concerns about privacy and data security.
• Accuracy and transparency. Understanding how AI analyzes data and creates content can be challenging as its outputs are sometimes ambiguous, incorrect or uncited. Nonprofits, for instance, are questioning grantmaking decisions as a growing number of funders evaluate grant proposals with AI.
• Job security. Funders and nonprofits alike share the global labor market’s fear that AI is replacing human workers.

Given these concerns, responsible use of AI requires thoughtful oversight and governance, clear expectations and a collective understanding of how and when the technology is used.

From Principles to Practice: Governing AI in Philanthropy
Recognizing AI’s risks doesn’t mean stepping back from its innovation, though. Rather, it underscores the importance of approaching the technology with clear guardrails.

Increasingly, foundations and donor-advised fund sponsors are defining how AI may be used – and where it may not – in their own organizations. Some applications are simply helpful timesavers, such as drafting summaries, analyzing trends or organizing information. Others carry greater responsibility, such as tools that influence funding recommendations or strategic decisions. Distinguishing between these uses helps ensure that human judgment is present and held accountable.

Human oversight is especially important in funding decisions. While AI can synthesize information and reveal insights, the expertise and perspectives of program staff and sector leaders add depth and breadth. Setting clear expectations around review processes helps prevent over-reliance on the technology.

Governance in philanthropy also extends to selecting and monitoring AI vendors. For funders, it’s important to understand how AI platforms handle data retention, whether submitted information is used to train new AI models, and who owns the intellectual property created through AI-assisted processes. A brief review of vendor terms and internal workflows can go a long way toward reducing unintended exposure.

Sound governance and oversight do more than mitigate risk. They reinforce accountability, strengthen transparency with grantees and stakeholders, and build trust.

AI, Data and Liability: What Funders Need to Know
Many AI tools process sensitive donor, beneficiary and grantee information, which can have implications for data confidentiality, intellectual property and decision-making accountability.

Before using these tools, funders are considering:

• What information is being entered into AI systems – and how is it stored and protected?
• Do vendor agreements clearly address data use, confidentiality and training for new AI models?
• Who owns AI-generated outputs created on behalf of the foundation or nonprofit partner?
• Where is human judgment required before our funding decisions are finalized?

Establishing clarity on these points helps ensure that AI enhances operational effectiveness while maintaining the trust of donors, grantees and the communities served. In our next edition in this series, we’ll be evaluating AI’s impact within philanthropy, helping others ensure responsible AI, and what funders can do.

Want to learn more about the ways we support the philanthropic sector?
Schedule a call or reach us at 800-839-0054. Together, let’s #begiving.