AI Backlash: Why People Are Protesting Against AI

The AI backlash is a set of separate objections that happen to share a target. Writers and actors want control over their work and likeness, residents near data centers worry about power bills and land, parents worry about chatbots and children, and some researchers want development slowed down. Polls show concern is widespread, but street protests have stayed small, and the concrete wins so far came from contracts, courts and local planning boards. I am a final-year Computer Science student and web developer (see about me), not a policy analyst. This post reports what dated sources say, separates fact from inference, and keeps my own view to one short section near the end.

What is behind the AI backlash?

"Backlash" is a loose label. In practice the objections play out in four places: union negotiating tables and courtrooms (creators), county planning meetings (communities), legislatures and regulators (children, privacy), and open letters (safety researchers). Marches get the headlines, yet they are the smallest part. Take the London march of 28 February 2026, organised by PauseAI and Pull the Plug, who called it the largest protest of its kind. A reporter from MIT Technology Review who walked with it described "a couple of hundred" people. The concerns on show ranged from online junk content and unemployment to extinction risk and AI near nuclear weapons, and the reporter noted that participants doubted companies would respond to pressure. A crowd with that many reasons is hard to turn into one demand. The data center fights, covered below, look very different: dozens of projects, specific officials, specific votes.

What do opinion polls say about AI?

Pew Research Center asked US adults the same question over several years. In a survey of 3,488 adults fielded 22 to 28 June 2026, 52 percent said they were more concerned than excited about AI in daily life, up from 37 percent in 2021. Among adults under 30 it was 55 percent, the first time a majority of that group said so. In the same release, 71 percent expected AI to mean fewer jobs in the US over the next 20 years, and 5 percent expected more. An earlier Pew survey of 5,119 adults (17 to 23 February 2026, published June 2026) asked broader questions. Forty percent expected AI to have a negative effect on society over 20 years, 16 percent a positive one, and 31 percent an equal mix. Sixty-three percent said AI is advancing too quickly. Sixty-seven percent had little or no confidence that the US government can regulate it effectively, and 59 percent had little or no confidence that US companies will develop it responsibly. The mood differs by country. Pew's 25-country survey from 2025 found a median of 34 percent more concerned than excited, 42 percent equally both, and 16 percent more excited. The US and Italy sat near half, South Korea at 16 percent. India had the highest share trusting its own country to regulate AI, at 89 percent. I could not find a reliable Pakistani poll on this, so I will not guess. One limit applies to all of it. "More concerned than excited" measures a mood. It does not tell you who will turn up to a protest or a council meeting.

Why are creative workers and publishers fighting AI companies?

Hollywood produced the clearest early case. The Writers Guild strike of 2023 lasted 146 days. In the deal, material generated by AI does not count as literary source material, and studios cannot require writers to use AI software, though writers may use it if they choose. The actors' union struck for 118 days and won rules requiring consent before a studio makes a digital replica of a performer, plus notice before synthetic performers are created. SAG-AFTRA members ratified a new four-year contract in June 2026 with more than 90 percent approval (about 19 percent of eligible members voted, according to Fortune). The union says its terms require AI performers to bring "significant additional value" over a live actor before producers can use them. The legal fights run on a different clock. In July 2026 a US court gave final approval to a 1.5 billion dollar settlement between authors, publishers and Anthropic, according to the Association of American Publishers. The case concerned nearly half a million books the company downloaded from pirate sites. That is a deal about how copies were obtained. It does not settle whether training on lawfully obtained works is allowed. That wider question is still open. In Getty Images v Stability AI, the UK High Court ruled in November 2025 that Getty largely lost on copyright (the judge held the model does not store copies) and won only a narrow trademark point about watermarks. The New York Times case against OpenAI and Microsoft, filed in 2023, was still in pretrial stages in the sources I found for mid-2026, and I could not confirm a trial date. The industry's strongest argument is that learning from existing work is how both people and software develop skills, and that a trained model is not a copy of its inputs. The creators' strongest reply is that a person reading a book bought it, and a model's output can compete directly with the people whose work shaped it. Courts have so far split the difference case by case.

Why are communities blocking data centers?

This is where opposition has turned into measurable results. Data Center Watch, a tracker run by the research firm 10a Labs, counted at least 75 projects worth about 130 billion dollars blocked or delayed in the first quarter of 2026, and at least 45 worth nearly 68 billion in the second quarter (Bisnow, 21 September 2026). The tracker says local opposition groups more than doubled from their 2025 level, and Loudoun County in Virginia directed staff to prepare a 12-month pause on new applications. The article does not define "blocked" versus "delayed", so treat the totals as an indicator, not a count of cancellations. Electricity prices are a big part of the anger. In July 2026 PJM, the grid operator for a large part of the eastern US, held a capacity auction for 2028 and 2029 with about 16.4 billion dollars in payments. Its independent market monitor estimated that about 6.3 billion of that came from data center demand, and argued that the cost lands on all customers, not on the data centers. PJM's own position is that the system needs more generation to meet growing demand. The energy agency IEA gives the counterweight. It estimated that data centres used about 1.5 percent of the world's electricity in 2024, and expects them to account for around one-tenth of global demand growth to 2030. In the US the picture is sharper: the IEA expects data centres to make up almost half of electricity demand growth. Both facts hold. Globally the share is small, and in a few grids it is large enough to move prices. The industry has responded. On 13 January 2026 Microsoft published five community commitments, the first being that it would "pay our way to ensure our datacenters don't increase your electricity prices". In March 2026 the White House announced a Ratepayer Protection Pledge that Amazon, Google, Meta, Microsoft, OpenAI, Oracle and xAI reportedly signed. Coverage describes the pledge as voluntary, and whether it holds is an open question.

Are workers and unions pushing back too?

Mostly through bargaining. In October 2024 the International Longshoremen's Association struck US East and Gulf Coast ports for three days, and members ratified a six-year contract in February 2025 running to September 2030. The union says it won full protection against automation, while AP and Bloomberg described a compromise that allows technology that does not affect jobs. That dispute was about automation in general, not AI specifically, but it is the same argument. For white-collar work the evidence is thinner. Pew's jobs numbers show what people fear, and fear is not the same as measured loss. The post on big tech layoffs and AI looks at what companies actually said about their cuts, and the one on whether AI will replace programmers covers the entry-level evidence for developers.

What are parents, privacy campaigners and safety researchers asking for?

Child safety moved fastest in the courts and statehouses. In January 2026 Character.AI and Google agreed in principle to settle lawsuits from families who said chatbot use contributed to teen suicides and self-harm. The terms are confidential, and no confirmation of final court approval turned up in my search. Character.AI had announced in October 2025 that under-18 users would lose open-ended chats. In September 2025 the US Federal Trade Commission sent orders to seven companies, including Alphabet, Meta, OpenAI and Snap, asking how they test and monitor companion chatbots for harm to children. California's SB 243, signed on 13 October 2025 and effective 1 January 2026, set rules for companion chatbots, including safeguards for minors. Facial recognition is an older fight with real rules. The EU AI Act has banned untargeted scraping of images to build face databases, and real-time remote biometric identification by police in public spaces (with narrow exceptions), since February 2025. The Dutch data protection authority fined Clearview AI 30.5 million euros in 2024 for building such a database. There is more on surveillance in is privacy dead. Safety researchers have used open letters. On 22 March 2023 the Future of Life Institute called on all AI labs to pause for at least six months the training of systems more powerful than GPT-4. Its page now shows more than 31,000 signatures. No major lab paused, as far as the reporting I found shows, and larger models have shipped since. A follow-up Statement on Superintelligence, released in October 2025, asked for a prohibition on developing superintelligence until there is broad scientific consensus that it can be done safely and strong public buy-in. When I checked, its page showed 78,423 signatures.

What have the protests actually achieved?

The record is uneven, and it is clearest where the demand is specific and someone has the authority to grant it. Contracts: consent and notice rules for digital replicas, limits on AI-written source material, and automation language in a dock contract. Courts: a 1.5 billion dollar settlement over pirated books, with the training question undecided. Local government: paused or blocked data centers and new cost-sharing rules, plus voluntary pledges from the companies. Regulators: child-safety scrutiny and a ban on some biometric uses in the EU. What has not happened: no pause on frontier training, no national limit on building data centers, and no settled ruling that training on copyrighted work needs a licence. That is a reading of the sources above, not a finding any of them states.

Text

CONCERN        STRONGEST EVIDENCE (as of November 2026)               BIGGEST UNCERTAINTY
Jobs           Pew: 71% of US adults expect fewer jobs (Aug 2026);    Economy-wide loss is hard to
               entry-level hiring studies                              separate from other causes
Copyright      1.5B dollar settlement over pirated books (Jul 2026);  Is training itself lawful?
               mixed court rulings                                     Cases still pending
Environment    IEA: 1.5% of world electricity (2024), but nearly      Local effects vary by grid,
               half of US demand growth to 2030                        water and cooling design
Safety         FLI letters; Pew: 63% say AI advances too quickly      How to measure and compare
               (Feb 2026)                                              risks that have not happened
Privacy        EU bans on some biometric uses; Clearview fine         Enforcement across borders
Power          PJM monitor: about 6.3B of a 16.4B auction tied to     Who pays, and whether voluntary
               data centers; pledges by Microsoft and others           pledges are kept

Is some of the opposition misdirected, and is some of it under-heard?

Both can be true, and the honest answer differs by topic. Arguments that some opposition is misdirected: AI is a small slice of world electricity today, per the IEA, and a data center also brings tax revenue and construction work. Some claims at protests lack support. The MIT Technology Review reporter flagged one claim about harmful infrasound from data centers as conspiracy-adjacent. Arguments that some concerns are under-heard: the PJM numbers suggest ordinary customers can pay for infrastructure built for a few large users, which is a fair thing to object to whatever one thinks of AI. Parents turned to the courts before chatbot rules existed. Creative workers argue that they were asked to compete with systems built from their own output, often with no say or payment. And on safety, a letter signed by tens of thousands of people did not change what labs do, which says something about how little influence the public has over research decisions.

What does the backlash mean for people who build software?

Developers sit on both sides of this. We use these tools and we build products on them, and the people objecting are often our clients' customers. A few practical points follow. Treat data sources as a legal and reputational question as well as a technical one, and know where training or reference data came from. Be clear with users when they are talking to a machine, since chatbot rules for minors are tightening. Plan for cost and energy questions if you run AI features at scale. And keep a human decision in places where a wrong answer hurts someone. In my final-year project, GoCart, the AI features are still in development, and these are the questions I try to ask before adding one. For the wider picture of benefits and risks, see the future of AI.

My take

I am a student developer who follows this topic, not an economist or lawyer, so weigh this lightly. The evidence looks strongest on specific, measurable harms and rules: who pays for grid upgrades, whether authors are paid when books are copied, what a chatbot may say to a teenager. These are old kinds of arguments about costs and consent, and institutions know how to handle them. That is why the wins cluster there. The broad claims look weaker, in both directions. "AI will take most jobs" runs ahead of what the layoff and hiring data show so far, and "the backlash is just fear of change" ignores real bills, real lawsuits and real settlements. The sources I read do not support either slogan. What would change this view: clear evidence that firms are cutting staff because of AI at scale, a court ruling that settles the training question one way, or data on whether voluntary power pledges actually keep household bills flat. The reading of small marches would also change if a movement began winning elections or legislation instead of staying at the edges. If you want to judge a claim yourself, check its date, its sample size and who produced the number.

Want to talk about this?

I am a full stack developer in Rahim Yar Khan, Pakistan, a final-year Computer Science student who works in web development at Hadi Dev Lab. If you want to discuss a project, or just compare notes on the sources above, you can message me on WhatsApp or by email. You can also see my work. Cover photo from Unsplash.