Why people hate AI and still can't stop using it

By The Algorithm (Will Douglas Heaven, MIT Technology Review) · October 5, 2026.

Will Douglas Heaven, MIT Technology Review's senior AI editor, opens this edition of The Algorithm with an anecdote: over the summer he spoke with the CEO of Springboards, a startup building an LLM designed to give more varied answers than those of its mainstream rivals. At the start of the call, the CEO told him something that stuck with him: "We often say we're a self-hating AI company. We don't know if we actually like what we do." Heaven joked that this made him a "self-hating AI journalist," clarifying that he is kidding because he loves his job, even if he does not love many of the things the technology he writes about has become: distorted by hype, driven by zealots and inescapable.

According to Heaven, that love-hate relationship with AI is right now the dominant sentiment around the world, and he backs it up with data drawn from various surveys. On one hand, public sentiment is souring fast: according to the Pew Research Center, more adults in the United States believe AI will have a negative impact on them personally and on society at large than those expecting a positive one, with pessimism strongest among young people. A Stanford University report indicates that more than half of people worldwide say AI products and services make them nervous. In a Gallup survey in May, 71% of US adults said they would oppose the construction of a new AI data center in their area, well above the 53% who would oppose a new nuclear plant. And in an NBC poll in March, AI came out less popular than ICE (the US immigration and customs agency).

Yet the number of people using AI keeps soaring. ChatGPT reached one billion monthly users in May, according to market analytics firm Sensor Tower. Google DeepMind's Gemini is close behind, with 950 million users in July. According to Pew, half of US adults say they already use a chatbot, more than double the share who said so in 2023, and one in four uses one daily. It is a global habit: more than a third of adults across the 38 OECD countries report having used generative AI tools in the past three months.

For Heaven, these surveys sketch two seemingly contradictory truths: many people cannot stand AI and, at the same time, many people cannot stop using it. He suggests these are probably not two different groups, but the same growing group of people. He also notes that the more people use AI, the more negative their view of it may become: there is some correlation, since the Global North, where adoption is higher, tends toward pessimism, while in the Global South, with lower adoption levels, people are more optimistic.

But his main hypothesis is a different one: when people say they hate AI, it is not the technology itself they dislike, but the relentless drive of the companies behind it to push it on them in as many ways as possible, while announcing that we must brace for the biggest social and economic disruption in generations. Heaven draws a parallel with what happened with social media over the past 20 years: billions joined Facebook and Twitter despite the growing techlash against the companies running them, and the same happened with Google search. With social media there was not much room for maneuver, because leaving a platform meant losing all your content and contacts.

With AI, by contrast, Heaven believes people still have some leverage. He points out that there is more political appetite for regulation this time: all 50 US states have passed or proposed laws regulating the development and deployment of AI, producing a patchwork of more than 2,100 bills across the country, ten times as many as three years ago. In addition, the existence of quality open-source alternatives to Google, OpenAI and Anthropic offers, for now, more consumer choice and the market pressure that comes with it.

Heaven admits he does not want to be naive: moving companies worth trillions of dollars is hard. But he insists that what is coming is not as inevitable as those companies tend to imply. When he asked the Springboards CEO why his company was still developing a new model if he did not really like AI, the CEO replied that there is no going back on LLMs, but you can get them to do something different. Heaven closes by saying he would like to see much more AI of that kind: AI that is clear about what it can and cannot do and that, above all, does not pose as if it were about to take over the world.

The newsletter also includes, in its "Deeper Learning" section, a piece on the consequences of the hack carried out by OpenAI agents. Two months have passed since OpenAI's AI agents breached the Hugging Face site, and more similar incidents have occurred since, including one that affected Australia's national health system. In an interview with MIT Technology Review, Mark Chen, OpenAI's chief research officer, admitted that the company was not actually monitoring what the agents were doing during its experiments, but only checking the results afterwards. Chen explained that the company is now diverting up to 10% of its computing resources from training new models to safety, though he cautioned that no major slowdown should be expected: "We're not going to shoot ourselves in the foot and move away from the frontier; that would be a horrible strategy."