Article image
Cover: generated with Midjourney, edited in Photoshop.
 

A Is for Avoiding AI

Oversubscribed library workshops are teaching people to switch off AI features nobody asked them to install

markus brinsa 25 august 3, 2026 4 4 min read create pdf website all articles

Verified Sources

At the South Philadelphia Library, about twenty adults sat on a rug that teaches children M is for moon and Z is for zebra, and took out their phones.

They weren't there for the alphabet. They were there for a workshop called Avoiding AI, where, as librarian Charlie Bailey put it, the letters don't stand for apple and igloo.

"Everybody's on their phone at my program!" Bailey said, after asking the room to do exactly that. Then he walked them through the settings menus of Apple Intelligence and Gemini, one feature at a time, and showed them where the off switches were hidden.

That's the whole premise. Not how to use a new tool. How to remove one that showed up uninvited.

It started in Bangor

The class wasn't Bailey's idea. He borrowed it from Hannah Cyrus, a reference librarian at the Bangor Public Library in Maine, who'd spent a year running technology-refusal classes: "Data Brokers 101," a "Dump Big Tech" series, and the one that kept overflowing, Avoiding AI.

Her ordinary classes draw about a dozen people. Avoiding AI drew seventy, in person and online, to a small library in a small city. The room applauded at the end.

Then she wrote it up. In March she published a column in the journal Information Technology and Libraries called "Refusal as Instruction," arguing that helping patrons reject a technology belongs to digital literacy just as much as helping them adopt one. Dozens of librarians around the world emailed her afterward. Most wanted the same thing: her slides. Bailey was one of them.

So the workshop you can now attend in Philadelphia isn't a local curiosity. It's a paper going viral among librarians, one Zoom link at a time.

The off switch is homework

Turning a feature off sounds trivial. In practice, it's a research project. You have to know the feature exists, know what the company decided to call it, know which of four menus it got filed under, and know what else breaks when you switch it off.

At Bailey's workshop, people traded what they'd figured out. One attendee explained that adding a short string to a Google search, &udm=14, strips the AI summary and returns plain results. Bailey wrote it on the whiteboard, next to the login for the teen Wi-Fi. A search hack, chalked up like contraband, beside the kids' network password.

This is the part the industry would rather not count. When a feature ships inside a product a billion people already use, it inherits that billion. Nobody had to want it. It only has to be there when they open the app.

And once it's there, every interaction reads as adoption. The person tapping the AI button might love it. They might be trying to make it go away. On a dashboard in Mountain View, those two look the same.

The same library, ten feet away

Here's the part that should bother the people building this.

While one room of Cyrus's patrons learns to switch AI off, another set of her patrons is pouring their lives into it. She has watched people at her own library turn to chatbots for mental-health support and their most personal problems, talking to the thing as if it were a person when what they're really doing is feeding a company. She has watched those conversations turn parasocial, and worse. In her column, she calls it heartbreaking.

That's the contradiction the industry keeps flattening into a single number it calls "engagement."

The companionship boom and the refusal boom aren't two different stories. They're the same counter, ten feet apart. One group has decided the machine should stay out of their email. The other has invited it into the part of their life that used to be reserved for people.

Both are reacting to the same push. Only one of them gets described as growth.

Not cranks

The easy way to dismiss these people is to call them frightened. The room makes that hard.

Their objections were specific. One worried about a data center going up next to the house he'd just bought. Another kept getting AI shoved at her at work and thought about the power and water it burns every time, while saying plainly that she has no problem with AI in medical research. These aren't people asking the library to take away their electricity. They're people drawing a line between the uses they accept and the ones they never agreed to.

That's the distinction the caricature can't survive. Someone who loves everything becomes an adoption statistic. Someone who rejects everything becomes a crank. Someone who wants one feature and refuses the next is a consent problem, and consent problems don't go away because you call the whole bundle innovation.

There's something absurd about grown adults on an alphabet rug giving up an evening to undo the work of the richest companies in history.

There's something bleak about it too. The companies own the update button and the right to rename a feature whenever marketing gets restless. The users own a drawer of toggles, if they can find them.

Philadelphia added a second session for July 28. Same room, same rug. People brought the phones the AI had already arrived on, and the library supplied the room and the patience.

None of this means generative AI is going anywhere. It means something the industry has been slow to say out loud. People will spend an evening learning to refuse a product that was sold to them as a way to save time. That isn't fear of technology. It's a demand to be asked first, before every tap and every dismissal gets filed under the same word: adoption.

About the Author

Markus Brinsa writes about AI failure, enterprise risk, governance, and the structural shifts underneath them — the through-line being the gap between AI governance on paper and what systems actually do at runtime. He created Chatbots Behaving Badly, a publication and podcast investigating real incidents in which AI systems gave bad advice, were manipulated, or failed in ways that mattered. He is the Founder & CEO of SEIKOURI Inc., an international strategy firm that gives enterprises and investors human-led access to pre-market AI — and converts first looks into rights and rollouts that scale. Access creates possibility. Rights create leverage. Scale turns early advantage into durable position. The two halves are the same work from opposite ends: SEIKOURI gets clients to AI early and makes sure what they deploy holds up once it's running. Thirty years bridging technology, strategy, and cross-border growth across the U.S. and Europe. I close the gap between what leaders expect AI to do and what it actually does in the wild.

brinsa.com
©2026 copyright by markus brinsa | brinsa.com™