Artificial intelligence can be a game-changer for nonprofit employees. While most people are using AI in their day-to-day work, they (rightfully!) worry about things like data privacy, avoiding bias, and intellectual property theft. That’s why Neon One has carefully built an AI governance policy that dictates how AI fits into our products.
Do you use AI tools at your nonprofit? If you do, you might have some concerns about the hidden impact of using that technology.
GivingTuesday’s 2024 AI Readiness Report found that 68% of surveyed nonprofit staff had tried using AI in their work, and that percentage has undoubtedly gone up since then. But, even though a majority of people were using AI tools, around half were worried about data breaches, bias, and intellectual property theft.
As nonprofit tech providers, we’re concerned about similar things. Here is, in plain language (that was not generated by AI!), how Neon One approaches artificial intelligence in our platform.
How Does Neon One Approach AI?
Just like many people, we use AI to make our work faster and more efficient. Some of our engineers recently used it, for example, to create a quick prototype of a solution to a particular customer’s problem. Once they had a prototype, the rest of the engineering team took the concept, built it themselves, and got it to the customer in record time.
While we each use our own AI tools for different parts of our jobs, we know that people like you use AI for your job, too.
We know that AI is rapidly changing the way people work because we see it in our own jobs every day. We also know that AI tools built specifically for nonprofits need to be developed with a special level of care.
Your work depends entirely on making real people feel connected to your organization, your work, and the people you serve. Yes, AI can help you do your work faster. But faster work doesn’t matter if the end result falls flat with the people in your community.
Neon One approaches AI with the perspective that any tool we build should enhance your ability to build those connections.
No tool we build will ever make those connections for you, and they’ll never replace the human-to-human interactions that result in outcomes like better fundraising and higher donor retention rates. Our job is to make your job easier, not to make your job unnecessary.
Our Core Approach
Have you heard of Fundraising.AI? It’s a collective of people (read: it’s not affiliated with any particular company) who are working together to help guide the nonprofit sector’s response to AI. Their framework for ethical AI usage in fundraising is the cornerstone of our own internal policies.
You can learn more about their framework here, but here are the three key principles that we have taken on:
- Privacy and security: We protect personal data with strong security measures and comply with all data protection regulations.
- Data ethics and inclusiveness: We ensure that our customers’ data is collected, stored, and used in a way that keeps nonprofits and their donors safe. (More on this below!)
- Accountability and transparency: It’s our responsibility to clearly communicate how our platform uses AI and to create tools that follow ethical standards.
That’s a very (very) high-level overview of our general approach. But what does this look like in practice?
How We Approach Generative AI
If you’re not familiar with this phrase, “generative AI” describes Large Language Model (LLM) tools that create content like text, images, and video. If you’ve used something like ChatGPT or Gemini to create a rough draft of an email, talk to a chatbot on a website, or summarize a longer piece of content, you’ve used generative AI.
Generative AI tools learn from very large data sets. That’s fine in some instances, but it can also cause a lot of problems if those data sets include sensitive information.
That’s why Neon One strictly prohibits training those tools using customer data.
Your data will never be used to train an LLM or other generative tool. If you use a tool like Gen (our nonprofit assistant), that tool will only use data from your instance, and it will only do so when you ask it to.
If you want to get nerdy, here are two related technical notes (if you don’t want to get nerdy, just skip to the next section):
- We use Retrieval-Augmented Generation (RAG). We’ll retrieve relevant data from your secure system and inject it into the model’s context window only at the time of the request. This ensures that no data is ever used for LLM training. The data is removed from the model’s memory when the request ends.
- We only use LLM providers that offer a contractually binding zero-retention policy. No inputs or outputs will ever be logged by providers for their own model training.
Generative AI is only one kind of artificial intelligence. Predictive AI is also an increasingly valuable tool for nonprofits, which is why we have separate rules for it.
How We Approach Predictive AI
As the name implies, predictive AI uses machine learning and other techniques to predict what may happen in the future. Have you ever ordered something from Amazon and then gotten suggestions for related products? Or have you ever noticed that Uber prices surged right after a huge concert let out? You’ve experienced predictive AI in action.
Predictive AI can be incredibly valuable for nonprofits. But, to quote Spider-Man’s dearly departed Uncle Ben, “With great power comes great responsibility.”
We take that responsibility very seriously.
There are a lot of very technical details that go into how we create predictive tools that are both useful and safe for nonprofits and their supporters. But we’ll distill it into two key points:
- We have stringent safeguards in place to ensure that any predictive modeling is done with fairness and without bias.
- We avoid black boxes. Take The Generosity Indicator, for example: We’ll never show you someone’s score without giving you insight into how we calculated it.
These policies are front and center any time we decide how AI will work in the Neon One platform.
How We Approach Data Sovereignty
This is important for you to know: You have absolute control over whether or not AI tools are present in your platform at all.
Every Neon One account has an admin setting that allows or disallows data from your instance to flow into AI services. You retain control over your data. If you opt out, those tools are unavailable and will not be able to touch your data.
Before we even begin to develop any AI-powered tool or feature, we make sure that it will come with oversight and guardrails. The more powerful the tool, the more stringent the oversight.
Second, all AI content that appears on our platform is clearly denoted as being AI-generated. You will never have to second-guess whether or not you’re dealing with generative AI.
Sum It Up
Nonprofit technology, including AI, should help people build real relationships with each other. But AI is powerful, and that power means it must be used carefully.
That’s why Neon One is very deliberate in the way we approach incorporating AI into our platform. We will always prioritize people’s privacy, and we will always be transparent about when and how AI is built into Neon One.
We have strict guardrails in place that help us make sure that everything we build, whether it’s a simple generative feature or a powerful predictive one, is built with you and your donors in mind.
