AI should help people do more good

AI is going to be part of how work gets done, whether we like it or not. The real question is who controls it and what it’s used for. Here is how we think about it, and the rules we follow with every client.

Where we stand

We don’t think AI is magic, and we don’t think it’s evil. It’s a tool, like the printing press or the internet. A model is a set of data and a computer program, and people decide how it gets used.

Our view comes from Soapbox Technology, the company behind this service. Soapbox builds open-source software for activists, journalists and community groups around the world, and it has published an ethics pledge and a statement on open AI. We bring the same values to the nonprofits we work with.

When AI is good

  • It takes boring, repetitive work off people’s plates so they can spend time with clients, donors and volunteers.
  • It helps a small team do work that used to need a bigger budget, like translations, reports and grant drafts.
  • It lowers barriers for people without technical training, so more of your staff can build and fix things themselves.
  • It runs on open models and open tools that anyone can inspect, and that no single company can take away.
  • A person reviews the output and stays responsible for the result.

When AI is harmful

  • It is used to deceive people, such as fake photos of clients or invented statistics in a grant report.
  • It makes decisions about people, like who gets services or who gets hired, with no human judgment involved.
  • It collects sensitive data about the people you serve and sends it somewhere you can’t see or control.
  • It is a closed black box owned by a company that can change the price, the rules or the terms at any time.
  • It is used as a reason to cut the staff who hold relationships with your community.

The world we want AI to help build

Picture a food bank where the operations manager spends Friday afternoon with volunteers instead of rebuilding a spreadsheet. A three-person advocacy group that can read every bill in a legislative session. A literacy program that sends families updates in Spanish and Vietnamese without hiring two translators. An executive director who writes the grant herself, but starts from a solid draft instead of a blank page.

That’s the use of AI we care about. Small organizations doing the work of communities, with the paperwork shrunk down so people have time for each other.

It only works if the tools stay open and affordable. If a handful of companies control every model and can raise prices or change the rules overnight, nonprofits end up dependent on them all over again. That’s why we favor open-source models and tools whenever they do the job, and why we document everything so you can switch providers or consultants later.

What “open” means to us

We follow the Open Source Initiative’s definition of open-source AI: you should be free to use, study, change and share the system. In practice that means:

  • Open code that anyone can read and improve
  • Open model weights, so the model can be run by anyone, anywhere
  • Enough information about training data to understand how the model was made
  • Access through services like OpenRouter that let you choose your provider and switch without starting over

Hosting a model costs money, so open doesn’t always mean free. It does mean nobody can lock you in.

Our commitments to every client

  • We tell you honestly when AI is the wrong tool for a job.
  • A person on your team approves anything that goes to a donor, client or funder.
  • We never use AI to decide who receives services or who gets hired.
  • We set up accounts that don’t train on your data, and we write down what information stays out of AI entirely.
  • We offer open-source models and tools as an option on every project.
  • We train your staff on ethics and security so the rules stick after we leave.
  • We document your setup so you aren’t dependent on us.
  • We never share or sell your data, or your clients’ data.

Common concerns we hear

Will AI take jobs at our organization?

At most small nonprofits the problem is too much work for too few people. We set AI up to give staff time back. We won’t help design a plan whose main goal is cutting the people who hold your relationships with the community.

What about bias?

AI models learn from the internet, and they repeat its stereotypes. That’s one reason AI should never rank or screen people, and why staff should read everything it writes about the communities you serve before it goes out.

What about energy use?

AI uses real electricity. Smaller open models often handle routine tasks well and use far less computing power, so we match the model to the job instead of defaulting to the biggest one.

Should we tell donors we use AI?

Follow any funder rules first. Beyond that, being open about it builds trust. Many organizations add a line to their policy or website explaining how they use AI and how they protect people’s information.

Put it in writing

Our free AI use policy template covers approved tools, protected information, human review, disclosure and security. Fill in the brackets and take it to your board.

Download the free template Ethics and security training

Texas Nonprofit AI Consulting is a service of Soapbox Technology LLC. Practical, ethical AI consulting for Texas nonprofits, colleges, schools and advocacy groups. Email alex@soapbox.pub.

In person within about an hour’s drive of greater Houston, and by video anywhere in Texas.