top of page
Our Policy on AI Tools

The short version: NGA CAN uses AI tools. We use them because we are a volunteer organization with limited money, limited time, and work that sometimes has to happen within hours. We also believe the AI industry needs substantially stronger regulation, including protections for artists, transparency around training data, and real limits on the energy, water, and infrastructure costs being pushed onto communities.

 

Four things hold this whole position together:

​

We use AI tools. We are transparent about it. Humans stay accountable for everything we publish. We want the industry regulated, and we organize for that regulation.

​

Everything below explains those four points. If you only read this far, you've got the gist.

WHO WE ARE, SO THE REST MAKES SENSE

NGA CAN is an all-volunteer progressive organizing collective serving Cherokee, Pickens, Bartow, Forsyth, and Cobb counties.

We run on essentially zero budget. Our work covers a lot of ground. We run ICE rapid response and know-your-rights education in English and Spanish for neighbors facing enforcement actions. We publish voter guides across five counties, interview candidates, and defend the right to vote and the democratic process itself. We run recurring mutual aid events where neighbors give and get what they need for free. We hold local officials accountable through public records, public comment, and plain old showing up. And we fight the data center gold rush in our own backyard, the kind of proposals that drop hyperscale facilities next to people's homes, sometimes over the written objection of a city's own planning staff, with no independent environmental study and no health review.

​

We are not AI boosters. We take no AI money. We believe this is a political emergency, and we use every lawful tool available to fight fascism, ICE overreach, and the corporate capture of local government. AI is one of those tools, and it helps on every one of those fronts.

​

Some NGA CAN members disagree with using AI at all. We do not require everyone in the organization to share the same view on every technology or ethical question. What we do require is adherence to our standards for accuracy, transparency, privacy, and human accountability, which are spelled out below.

WHAT WE USE AI FOR, AND WHAT WE DON'T

We use it to read and summarize long public documents like rezoning files and meeting agendas, to draft and format newsletters and social posts, to produce transactional graphics and flyers, and to move fast on rapid response communications.

 

We do not use it to replace paid work that we have the money to pay for, because we almost never have that money, and when we do, people come first. We do not use it as a substitute for human judgment about what to say or whether something is true.

OUR INTERNAL RULES FOR USING AI

These are the commitments a supporter should actually be able to hold us to. They matter more than whether a rally graphic could have been hand-lettered.

  • Humans verify every factual claim against primary sources before we publish it.

  • Humans make the final editorial decision on everything that goes out.

  • We do not fabricate quotes, sources, events, photographs, or evidence.

  • We do not use AI to impersonate real people or to generate fake constituent testimony.

  • We do not use AI-generated images to depict real people or real events in ways that could reasonably be mistaken for documentary photographs, and we disclose AI-generated imagery anywhere the distinction could mislead.

  • We do not use AI to replace paid labor when funding exists to pay for that work.

  • We correct our errors publicly when we make them.

  • We do not feed sensitive personal information into consumer AI tools. This one is not optional for us. We do ICE rapid response and know-your-rights work, and protecting the people who trust us with their information comes before any convenience a tool could offer.

THE APPARENT CONTRADICTION: FIGHTING DATA CENTERS WHILE USING AI

This is the objection we hear most. The short answer is that we do not oppose computation existing. We oppose speculative hyperscale facilities dropped next to people's homes with no noise standards, undisclosed water use, secret stipulation letters, tax giveaways, and utility costs socialized onto residents while the profits stay private.

​

Those are two different things.

​

Opposing a coal plant next to an elementary school does not require living without electricity. Opposing a reckless highway widening does not require never driving. Food justice organizers still eat food grown in the industrial food system, because the point was never personal purity, it was changing the system. You fight the thing that is harmful about the system while still living inside it, because there is no outside to live in.
 

We are not asking anyone to achieve personal purity inside systems they cannot opt out of. We are trying to change the systems themselves. Individual consumption choices are not an adequate substitute for regulating the institutions producing the harm. One volunteer not typing into a chatbot changes nothing about the buildout, which runs on hundreds of billions of dollars of corporate capital, not on a nonprofit's use of a writing tool. What actually stops a bad data center is organizing: forcing votes onto the public record, showing up to hearings, and winning real regulation and siting standards.

​

The philosopher Alexis Shotwell wrote a book about this trap, "Against Purity: Living Ethically in Compromised Times" (University of Minnesota Press, 2016). Her argument, roughly, is that there is no ethically pure position to organize from, and demanding one is a way to guarantee you never act. We agree. We would rather be effective than perfect.

THE HARM TO ARTISTS

When a model is trained to reproduce a specific living artist's style on command, when someone's life's work gets scraped without permission or payment, when the market gets flooded with cheap imitations that undercut the people who built the skill, that is real harm to real people, and it is happening now regardless of how the copyright cases come out. Artists are right to be angry about it.

​

Two things are true at once. The harm to working artists deserves a real response, through law, licensing, and compensation, all of which we support. And a volunteer group generating a rally graphic did not cause that harm, took no commission from anyone, and cannot fix it by abstaining. We support the fight for artist protections. We do not pretend that a nonprofit skipping a flyer is that fight.

​

On the narrower legal question, the courts have not handed anyone a clean answer. Federal judges have split: some have found that training on lawfully acquired material is transformative, one found piracy of the source material was not defensible and produced a very large settlement, and other cases are still moving through the system with no settled outcome. Anyone who tells you the copyright question is settled, in either direction, is not being straight with you. We think the current US framing of intellectual property is a legal and social construct worth arguing about, not a settled moral absolute, and we would rather see it resolved through legislation and licensing than pretend it already has been.

​

WHAT THE TECHNOLOGY DOES

It helps to know what is actually happening, because a lot of people have an understanding of the technology that is not quite right.
 

A generative model does not keep a folder of images and paste pieces together. It works by prediction. During training it is shown enormous amounts of text or images and learns the patterns in them: which word tends to follow another, what a sunset usually looks like, how a protest sign is laid out. After that training, when you give it a prompt, it generates something new one piece at a time, each step a best guess at what should come next based on those learned patterns. A text model is predicting the next chunk of text. An image model is predicting what a fitting image looks like. It is closer to how a person who has studied thousands of posters can sketch a new one than to a scrapbook of cut-up originals. Models can sometimes reproduce something close to a specific training example, which is a real and studied problem, but that is the exception, not how an everyday graphic gets made.

​

This is where data centers enter the picture, and where the two halves of this whole conversation connect. All of that pattern-learning and predicting runs on specialized computer chips, and those chips live in data centers. There are two very different costs. Training a model, the one-time process of learning the patterns, is incredibly energy intensive and happens on huge clusters of those chips. Using an already-trained model to answer a single prompt, which is what we do when we make a flyer, takes a small fraction of that, closer to the cost of a short web task than to anything dramatic. The specific numbers, and how they compare to everyday things like streaming, are in the energy section below.

​

So the data center fight and the AI question are the same physical story seen from two ends. We use the finished tool, which is the cheap end. The facilities being built across Georgia are sized for the expensive end, training and running these models at massive commercial scale, and that is the buildout we organize to regulate. One resident making a graphic is not what drives a new large-scale facility next to an apartment complex. Industry-scale demand is.

​

ENERGY AND WATER

A typical text query uses a relatively small amount of energy and water compared with the aggregate footprint of the infrastructure supporting AI. Image generation uses more, and estimates vary substantially depending on the model, hardware, and system. Those per-use numbers are not where the real problem lives.

​

The real problem is aggregate. Training is resource intensive, and it happens repeatedly across model generations and fine-tuning, not just once. Inference creates an ongoing footprint that grows with use. Put those together at industry scale and data center demand becomes a strain on grids and water systems. The International Energy Agency estimates global data centers used roughly 1.5 percent of world electricity in 2024 and projects that to roughly double by 2030. A US Department of Energy report prepared by Lawrence Berkeley National Laboratory found US data centers used about 4.4 percent of national electricity in 2023 and projected between 6.7 and 12 percent by 2028.

​

That aggregate load, concentrated in specific communities, is exactly what we organize to regulate. The relevant policy questions are the infrastructure and where and how it gets built: require energy and water disclosure, require efficient closed-loop cooling, regulate siting, and stop pretending a resident's individual query is the mechanism driving local demand. One volunteer abstaining has no meaningful effect on the aggregate buildout. A moratorium and  standards change it a lot.

THE TRUE TECH THREAT IS NOT OUR FLYER OR SOCIAL POST

If you are worried about technology and power, aim at the actual target. Journalist Gil Duran, in his book "The Nerd Reich: Silicon Valley Fascism and the War on Democracy," documents how a small circle of Silicon Valley billionaires came to see democracy itself as an obstacle and decided they should be governing in its place. As Duran put it in an interview on NPR's Fresh Air, this is a group who believe that in the 21st century technology will make democracy obsolete, and the breaking of public institutions is the goal, not an accident.

​

That is the tech fight that matters, and it is the same fight we are already in locally. Out-of-state capital buys up land. Captured officials wave it through. The public gets the noise, the water draw, and the power bill, while the profits leave town. A data center dropped next to homes with no standards and no accountability is that story in miniature. So when someone spends their energy policing whether a volunteer used a chatbot, they have the threat backwards. The danger was never neighbors using tools. It is a handful of the richest people alive building infrastructure, and buying government, that answers to no one. Read Duran's reporting at https://www.thenerdreich.com and decide for yourself where the real power grab is.

WHAT WE SUPPORT

We are not neutral on AI policy. We want it regulated. Specifically:

 

  • Transparency requirements for AI training data.

  • Compensation frameworks for artists and writers.

  • Mandatory energy and water disclosure for data centers.

  • Local zoning standards for data center noise, equipment, and screening.

  • Moratoriums on new data center approvals until real standards exist.

  • Independent environmental and health review before any approval.

  • Stronger federal AI regulation across the board.

 

That is a more demanding agenda than "personally don't use the chatbot," and it is one that could change conditions on the ground.

APPENDIX: QUICK ANSWERS TO RECURRING QUESTIONS

The comments we get most often, and our answers:

​

"You fight data centers but you use AI. Hypocrites." Opposing a badly sited industrial facility next to homes is not the same as refusing to use software. One volunteer's query does not build or stop a data center. The buildout runs on corporate billions. Regulation and siting standards are the fix, not personal abstention.

 

"You used AI to write this too?" Sometimes. It lets a handful of unpaid people keep pace with full-time developers and lawyers. You can think our facts are correct and still object to the tool on ethics, that is a fair thing to hold, but the facts stand on their own either way.

​

"A real activist would make the flyer by hand." We are volunteers with no design budget and hearings scheduled 48 hours before a vote. The choice was never a hand-lettered poster versus AI. It was AI versus the flyer not existing.

​

"AI steals from artists." The harm to artists is real and we support artist protections, compensation, and regulation. The narrower copyright question is unsettled in the courts. And our flyer displaced zero commissions, because we never had a budget to commission anyone.

​

"The energy and water use is killing the planet." Per use, a query is tiny, smaller than seconds of streaming. The real driver is aggregate corporate buildout, which is exactly what we organize to regulate through disclosure and siting standards.
 

"Data centers are the whole problem, and AI is why they're built." Data centers run your bank, hospital records, 911 dispatch, the grid, your email, your streaming, and the platform you typed that comment into. AI is accelerating the buildout, which is why we fight for standards. The fight was never "data centers or none." It is whether they get built responsibly or recklessly.

​

"So you support AI. Astounding." We support using tools that exist and regulating them heavily. Both can be true.

​

"Get the truth before you speak." Everything we post comes from public records, linked so you can check it. If something is wrong, name the specific thing and we will correct it publicly. We have done that before.

"By commenting here you're feeding AI too." True, and so is everyone else. Meta trains its own AI on public posts and comments, and US users cannot fully opt out. Nobody in the thread is standing outside the system, which is the whole point about purity versus effectiveness.

 

"You're why my power bill is going up." No, the data center industry is, and that is a real harm we organize against. Costs get socialized onto residents while profits stay private. A volunteer's flyer has nothing to do with your bill. The unregulated buildout has everything to do with it.

"Just be anti-AI then." Being anti-AI accomplishes nothing on its own. The technology is not going anywhere, and if everyone who wants regulation swore it off, the only people left using it would be the developers and operatives already deploying it at scale. We would rather keep the tool and win the regulation.
 

"No one likes AI. You will not reach anyone posting this. Gen Z hates AI." Gen Z is one of the heaviest AI-using generations there is. Plenty of people dislike AI art specifically and still use AI daily, both things are true. But we do not pick our tactics by vibes. We pick them by what reaches people and moves them to act, and a clear, fast, sourced graphic about a hearing reaches more neighbors than a blank page does. If a post actually flops, we will change it. "People online say they hate AI" is not the same as a flyer failing to do its job.

​

"This is lazy." Reading a 200-page rezoning file, cross-checking it against the staff analysis and the agenda, and turning it into something a neighbor can act on in the 48 hours before a vote is not lazy. The layout tool is the fast part. The work is the research, and the research is done by a human who can defend every line of it.

​

"Just hire a local artist. Support your community." We have no budget to hire anyone. We are volunteers. When there is money for paid work, people come first, that is in our rules. If an artist wants to send us something to use, we will gladly try to use it, but we are not going to pretend a commission existed to be lost when it never did.

​

"Why not just get art from volunteers?" We have tried it, and for a group our size it does not work. Coordinating volunteer artists means briefing someone, waiting on a draft, asking for changes, and waiting again, all on a timeline where a hearing can be 48 hours out and details change by the hour. It is a real pain to manage and it is never as fast or as flexible as the work demands. Some groups have the capacity to run a volunteer art team well. We do not. If an artist sends us something finished and ready to use, we will gladly try to use it. What we cannot do is build our rapid response around waiting on it.

​

"You lose all credibility using AI." Our credibility rests on whether our facts are right and our work holds up at the podium and in the public record. It has never once rested on which program set the type. Check the sources. That is the credibility test that matters.

​

"Real grassroots groups do not use corporate AI tools." Real grassroots groups use email, phones, Facebook, Google Docs, and printers, all made by large corporations, because the alternative is not organizing at all. We use the tools available to us and we fight to regulate the companies that make them. Purity has never freed anyone.

​

"If your message were good you would not need AI to make it look nice." The message is the facts about the data center. The graphic is how we get those facts in front of people who are scrolling. Good organizing has always used whatever made the message land, from mimeographed flyers to spray paint to social media. The tool is not the argument.

​

Sources
bottom of page