Avoiding AI Workshops Go Viral as Libraries Answer Demand for More Control

The techcrunch librarians story captured a conflict Big Tech has struggled to acknowledge: many people want less AI, not another reason to adopt it. Libraries in several American cities are answering that demand with workshops devoted to disabling unwanted artificial intelligence features.

These sessions reverse the familiar AI literacy model. Participants still learn how generative AI works, but the practical lesson concerns refusal. Librarians demonstrate settings, alternatives, and workarounds that let people choose which systems operate on their devices.

The response has exceeded attendance for ordinary computer classes. That matters because technology companies often describe AI adoption as an inevitable progression. The workshops expose a different reality: products can gain distribution through defaults even when users remain unconvinced.

Avoiding AI Became an Unexpected Library Hit

A modest technology class became a visible measure of public frustration with unwanted AI features.

Charlie Bailey, a librarian at the South Philadelphia Library, held an hour-long Avoiding AI workshop on June 29, 2026. The official event listing promised instructions for disabling AI tools in operating systems, applications, and popular websites.

About 20 adults attended the Philadelphia session. Bailey began with an overview of consumer AI systems before guiding participants through their phones and computers. The practical demonstrations included turning off Apple Intelligence and Google Gemini features.

That structure makes the workshop more substantial than an anti-technology gathering. Participants hear why someone might use an AI tool, why another person might abstain, and what control remains available. The objective is informed choice rather than universal rejection.

Bailey told TechCrunch that frustration with imposed tools inspired the program. People were encountering AI features they had not requested, often inside products they already used. Some settings were difficult to find, while other integrations offered only partial opt-outs.

The event’s online response was much larger than its classroom capacity. According to the original coverage, the library’s Instagram announcement received more than 2,000 likes and 220 shares. Bailey scheduled another session after registrations exceeded the first workshop’s available space.

Those figures stand out against the account’s usual engagement. TechCrunch reported that many library posts receive no more than a few dozen likes. A settings workshop had reached people far beyond its immediate neighborhood.

The concept originated with Hannah Cyrus, a digital media librarian at the Bangor Public Library in Maine. Cyrus developed her own Avoiding AI class after patrons repeatedly asked why software was composing emails, summarizing short messages, or inserting AI-generated answers into routine tasks.

Her ordinary introductory computer classes typically attracted around a dozen participants. Registration for her first Avoiding AI workshop reached 30 before she closed the room, created a waitlist, and added remote access. Roughly 70 people attended each of the first two sessions when online viewers were included.

Cyrus later published material about developing the workshop. Dozens of librarians from different locations contacted her for slides and guidance. She described that professional response as unlike anything generated by her previous instructional work.

The pattern then spread. Librarians in Philadelphia, Maine, and Boston adapted the idea for their communities. The Boston Public Library promoted a session covering unreliable output, privacy concerns, and ways to limit AI features.

That demand is the event behind the techcrunch librarians keyword. Yet the numbers matter less as a national adoption estimate than as a signal. Small public programs are discovering an audience that conventional AI training often overlooks.

This audience does not need another prompt-writing seminar. It wants a map of the exits.

Why AI Opt-Out Demand Is Rising Now

AI fatigue grows when a feature arrives through a software update before users have decided that they need it.

Consumer AI deployment has moved from optional websites into operating systems, search engines, email services, office software, and mobile applications. That distribution strategy places AI in existing workflows without requiring people to seek it out.

The distinction matters. Choosing to open a chatbot is different from finding generated text inside a search page or a writing suggestion inside an email. One is an intentional product decision. The other is a default experience controlled by the platform.

Defaults reduce the effort required for adoption, but they also obscure consent. A user might technically have an opt-out while lacking any reasonable path to discover it. Settings can move between updates, carry unfamiliar labels, or disable only part of a feature.

Avoiding AI workshops address that gap at the device level. They show people where controls exist, what those controls actually change, and which features cannot be fully removed. The process turns an abstract debate over AI ethics into a concrete settings menu.

Privacy is one source of concern. People want to know whether text, images, voice recordings, searches, or device activity leave their computers. They also want to understand whether a provider retains that information or uses it to improve a model.

Reliability is another. Generative AI produces new text by predicting likely sequences from learned patterns. It does not independently verify each statement before presenting an answer. That difference becomes important when software summarizes information or responds with apparent confidence.

Cyrus summarized the limitation bluntly in separate reporting: “Generative AI doesn’t generate information. It generates words.” The point is not that every output is false. It is that fluent presentation cannot substitute for checking evidence.

The Maine library movement has expanded beyond classes. Steven Brown, director of the Searsmont Town Library, began helping patrons remove or limit AI features on personal devices after attending a Cyrus webinar.

Brown framed the work around choice. A phone arrives with a particular configuration, but its owner does not always have to accept that configuration. Librarians can translate an unfamiliar technical decision into a manageable action.

The audience also includes people who object to the economic and cultural systems behind generative AI. Their concerns cover training data, creative labor, environmental costs, misinformation, surveillance, and corporate concentration.

A single workshop cannot settle those debates. It can, however, separate them from the narrower question of control. Someone should not need a complete theory of artificial intelligence before deciding whether an email program may rewrite a sentence.

This is why avoiding AI is becoming part of digital literacy. Literacy does not mean compulsory enthusiasm for every technology. It means understanding a system well enough to assess its benefits, limitations, and alternatives.

The techcrunch librarians story surfaced because the market’s adoption narrative has been incomplete. Product usage can rise while resentment rises with it. Distribution measures exposure, but it does not always measure approval.

TechCrunch Librarians Expose the Default Adoption Problem

The central conflict is not librarians against innovation. It is user agency against product design that treats AI exposure as consent.

Large technology companies have strong incentives to place AI across their existing products. Wider distribution generates usage, feedback, and opportunities to retain customers inside a platform. It also helps justify the extraordinary resources committed to AI infrastructure.

Users approach the same interface with different incentives. They want to complete a search, send a message, organize a document, or operate a phone. An AI feature might help, but it can also add uncertainty, clutter, or an unwanted data relationship.

That mismatch creates what the workshops make visible. Platforms optimize for feature adoption across millions of accounts. Librarians respond to individuals who cannot locate a switch.

An opt-out can exist without providing meaningful choice. The user must first recognize that a feature is active. They must understand its name, locate its controls, and evaluate the consequences of disabling it. Software updates can then alter the arrangement.

Some AI integrations are easier to disable than others. A device assistant might have a clear settings page. A search engine can place summaries into its main results experience without offering a permanent account-level removal option.

Workshop participants in Philadelphia shared their own workarounds. One attendee explained that adding &udm=14 to a Google search address can request a web-focused results view. Bailey wrote the string on a whiteboard for the room.

That scene shows why libraries are well positioned for this role. A workaround that circulates among technical users becomes usable by a broader public when someone demonstrates it, explains its limits, and answers questions.

The instruction also exposes how uneven user control can be. People should not need to memorize query parameters to obtain a simpler search experience. Yet a small piece of shared knowledge can restore some choice when a platform’s interface does not.

Apple, Google, Microsoft, Meta, and OpenAI are not identical opponents in this story. Their products, settings, and business models differ. Treating them as a single technical system would hide important distinctions.

They share a distribution logic, however. Each has an incentive to make AI a normal part of an existing digital environment. The public encounters that strategy as a collection of prompts, summaries, assistants, writing tools, and suggested actions.

The Avoiding AI movement challenges the assumption that normalization ends the debate. A feature can become common before it becomes trusted. It can also remain widely used because avoiding it requires effort.

This distinction should matter to product teams. Forced exposure can improve short-term engagement numbers while damaging long-term confidence. Users who cannot understand or reverse a setting may distrust adjacent features, including features they might otherwise value.

Clear controls would not necessarily reduce adoption. They could make adoption more meaningful. A person who understands what a feature does, activates it deliberately, and can later disable it provides a stronger signal than someone counted through a default.

The techcrunch librarians coverage therefore describes more than a viral workshop format. It reveals an unusual feedback channel for the technology industry. Libraries are collecting the support requests created by AI-first product design.

Libraries Are Turning Refusal Into Digital Literacy

Public libraries can discuss AI adoption without selling a product, protecting a quarterly target, or requiring technical expertise from participants.

Libraries have long helped people use computers, navigate online services, evaluate information, and protect basic privacy. Teaching someone to limit an AI assistant fits that history more naturally than the workshop’s provocative name suggests.

Their institutional position is important. A platform provider can explain its own controls, but it still benefits from continued usage. A library can begin with the patron’s objective, including the decision not to use the product.

That independence changes the classroom. Participants can ask whether a feature is necessary without first accepting the company’s framing. They can compare convenience against privacy, accuracy, creative control, or simple preference.

The workshop also creates a social setting for concerns that people often experience alone. A user might assume that everyone else welcomes an AI summary or knows how to disable it. A full registration list demonstrates that confusion and skepticism are shared.

Bailey described camaraderie among the Philadelphia participants. Attendees exchanged tips instead of receiving a one-directional lecture. The librarian provided structure, while the room contributed experience across devices and applications.

This peer element matters because AI controls are fragmented. No single instructor can anticipate every phone version, browser setting, account type, or software update. Community knowledge helps fill gaps, although every workaround still requires verification.

The wider public-engagement field supports this broader definition of literacy. A 2026 public agency framework from the Association of Science and Technology Centers identifies several roles for trusted learning institutions.

Those roles include explaining AI, helping people navigate safety risks, and supporting decisions about whether and how to adopt the technology. The framework treats agency as an educational outcome rather than assuming adoption is the desired result.

It also recognizes risks involving privacy, bias, misinformation, and scams. Museums and science centers are not libraries, but they share a public-facing mission. Both can translate technical systems for audiences outside professional technology circles.

Avoiding AI classes put that principle into an especially direct form. The curriculum can begin with how a system works, continue through its tradeoffs, and finish with actions chosen by the participant.

A Boston workshop listing follows that model. It promises an explanation of generative AI, discussion of reliability and privacy, and instructions for disabling features as much as possible.

The phrase “as much as possible” is essential. Total avoidance is difficult because machine-learning systems operate behind many services without visible chatbot interfaces. Spam filters, recommendation engines, fraud detection, and content ranking all involve automated models.

A useful workshop must therefore define its scope. Disabling a visible generative feature is not the same as removing every algorithm from a device. Switching products can also exchange one set of data practices for another.

Good instruction should acknowledge those limits. Otherwise, “Avoiding AI” becomes a promise no librarian can fulfill. The realistic objective is selective refusal backed by better understanding.

That model can serve both skeptics and willing users. Someone who learns how to turn a feature off also learns where it operates and what information it touches. The same knowledge supports a more deliberate decision to turn it back on.

What the Viral Numbers Do Not Prove

Packed workshops show unmet demand, but they do not establish that most consumers want to abandon AI products.

The attendance figures come from a small number of programs. Seventy participants in a hybrid library session is remarkable relative to a dozen-person computer class. It is not a representative national survey.

Social engagement has similar limits. More than 2,000 likes on a library post demonstrate that the topic traveled unusually far. They do not reveal how many people disabled a feature, maintained that choice, or rejected AI altogether.

The label itself can attract several audiences. Some participants oppose generative AI on principle. Others dislike one product, distrust automated summaries, or simply want fewer interruptions. A third group may attend because it wants to understand the technology before choosing.

Those motivations should not be collapsed into an anti-AI bloc. The strongest lesson concerns demand for control, not a verified collapse in consumer adoption.

The technical advice also has a short shelf life. Companies change menus, feature names, account requirements, and default behavior. Workshop materials that work in June can become incomplete after a major software update.

Librarians will need a maintenance process. Instructions should identify device versions, record testing dates, and distinguish permanent settings from temporary workarounds. Otherwise, the program could create the same confusion it aims to resolve.

Alternatives require scrutiny as well. A smaller browser or search provider can reduce exposure to generated answers while introducing different privacy, security, or accessibility tradeoffs. “Not from Big Tech” is not an automatic quality guarantee.

There is also a risk of framing AI literacy as avoidance alone. People need enough understanding to recognize manipulated media, scams, unreliable summaries, and automated decisions even if they never open a chatbot.

Refusal and recognition are different skills. A person can disable an assistant yet still encounter synthetic content through social media, search results, customer service, or messages from other people.

The best workshop model combines three layers. First, it explains the system in plain language. Second, it identifies meaningful risks and benefits. Third, it gives participants practical controls without dictating one acceptable choice.

Industry defenders can reasonably argue that AI features assist people with writing, accessibility, translation, research, and repetitive tasks. Some users actively want them. Removing or discouraging those tools across the board would substitute one rigid policy for another.

That is not what Bailey’s format appears to do. He reportedly explains possible uses before presenting reasons to abstain. The curriculum focuses on autonomy, not a blanket ban.

Big Tech can answer the criticism without abandoning AI. Companies can provide clearer notices, stable settings, genuine opt-in pathways, and documentation that ordinary users understand. They can also separate core functions from experimental generative additions.

The techcrunch librarians phenomenon will become more significant if workshops continue filling after the initial publicity fades. Repeat attendance, requests from new library systems, and sustained demand for updated materials would strengthen the case.

For now, the story is a sharp signal with a limited sample. It deserves attention precisely because technology companies rarely receive this kind of feedback through conventional product analytics.

Three Signals Will Show Whether Avoiding AI Lasts

The next stage depends on whether a viral workshop becomes durable public infrastructure, better product controls, or a short-lived protest format.

The first signal is replication. Cyrus received inquiries from dozens of librarians after sharing her approach, while programs appeared in Maine, Philadelphia, and Boston. The movement becomes more consequential if additional library systems add recurring classes rather than one-time events.

Recurring programs would show that demand survives press coverage. They would also require shared curricula, update schedules, and guidance for different devices. A maintained public resource would turn scattered tips into an accessible form of consumer support.

The second signal is the response from platform companies. Apple, Google, Microsoft, Meta, and other providers can reduce friction around AI controls without slowing their technical development.

Watch for settings that remain stable across updates, explain data handling clearly, and disable related features in one place. A genuine opt-in during setup would be an even stronger response.

If controls become easier to locate, the workshops will have influenced product expectations even without formal campaigning. If companies make AI harder to avoid, attendance could grow while distrust deepens.

The third signal is the content of future AI literacy programs. Libraries might divide their offerings into adoption, evaluation, safety, and refusal tracks. That would acknowledge that communities contain both eager users and committed skeptics.

A broader curriculum would weaken the claim that Avoiding AI is merely a backlash trend. It would establish refusal as one part of informed technology use, alongside verification, privacy, accessibility, and practical experimentation.

Readers should apply the same test to their own devices. Identify which AI features are active, what information they process, whether they provide value, and whether the available controls match your preferences.

Keeping a personal record of those choices can help when software updates change familiar settings. The same habit supports broader personal knowledge management, where useful information remains organized and retrievable under the user’s control.

The techcrunch librarians story ultimately asks a simple question that product dashboards cannot answer: when people use AI because opting out is difficult, what does adoption really mean?

Libraries are providing one response. Give people understandable information, a place to ask questions, and practical control. Then let them decide which technology deserves a place in their lives.