Posted on 24 mins read

The following is a terse list of harms caused by generative AI. From here on, the term “AI” is used as shorthand for generative AI, particularly LLMs like ChatGPT and Claude, as well as generative image and video models.

This post was written without any kind of AI assistance, even for research.

Browse by area of concern:

I care about the truth

AI has drowned the Internet in low-quality content

AI enables the creation of meaningless content with no original thought or research. There’s been an explosion of AI-generated web pages over the last few years, making it dramatically more difficult to find useful and accurate information online.

AI constantly generates harmful misinformation

AI has no means of distinguishing between truth and falsehood, and constantly says things that are subtly or even blatantly incorrect.

AI obscures primary sources of information

AI chatbots and search summaries have replaced traditional research for many people, obscuring primary sources and reducing the quality of information they find.

AI is eroding trust in our institutions

Use of AI is exacerbating the public’s declining trust in essential institutions, from local governments to civil courts.

Claims about AI’s abilities are highly exaggerated

AI’s effect on global productivity has been far more limited than advertised. Additionally, companies in multiple industries have repeatedly been caught making false, exaggerated, or unverifiable claims about the abilities of AI products and features.

AI promoters use fear and groupthink to force acceptance

Executives and employees alike feel pressured to speak positively about AI because doing otherwise would harm their careers.

I care about the economy

AI companies are financed by deceptive circular deals

Circular financing allows AI companies to claim high valuations and revenue forecasts without actually making money. For example: Nvidia invests $100M in OpenAI and OpenAI signs a deal promising to spend that $100M on AI hardware from Nvidia; Nvidia then claims $100M of expected revenue and OpenAI claims a $100M valuation on the shares it sold, without either of them creating any real market value.

AI has created a speculative bubble, risking economic disaster

AI spending and valuations have reached unrealistic highs, with many experts forecasting a crash that could mirror the dot-com bubble.

AI is an incredibly inefficient technology

Large Language Models scale poorly, getting slower and requiring more resources with each model size increase. The “economy of scale” does not appear to apply here.

AI has driven up prices for consumer electronics

AI development requires massive amounts of computer memory. The race to sign multi-year contracts and buy up inventory from chipmakers has caused electronics prices to rise, negatively affecting non-AI companies and everyday consumers.

AI is destroying the financial model of digital publishing

Digital media and news sites have lost a significant amount of web traffic since the advent of AI chatbots, causing serious financial difficulties.

AI often reduces productivity rather than increasing it

AI claims to make users faster and more productive, but often fails to live up to this promise.

AI is built on stolen labor

AI models are built on the consumption of millions of people’s work without their consent.

AI developers shamelessly disregard other people’s privacy, consent, and choices about how their work is used.

AI blatantly copies existing work

AI models have repeatedly been caught copying other people’s work without attribution or consent.

AI is used to create nonconsensual pornography

AI models have repeatedly been used to create pornographic images of real people, including children, without their consent.

AI powers mass surveillance

AI is used to support government surveillance of private citizens at a massive scale.

I care about preserving human skill and intelligence

AI use leads to cognitive decline

Habitual AI users lose mental agility, critical thinking skills, and expertise that they may have spent decades building.

AI allows students to pass classes without learning anything

Students who use AI to do their homework or write their essays are cheating themselves out of an education. The harm is not limited to them, however; some of them are future doctors, for example, and their lack of education will also harm their patients.

AI is a tool of mass de-skilling

People who use AI for a task gradually lose their ability to do the task without it. AI companies favor this, since it creates widespread dependence on their products.

AI reduces the incentive and ability to learn new skills

People use AI as a shortcut for tasks they would otherwise have to learn a skill in order to do. This reduces their ability to persist and limits opportunities for growth, learning, and specialization.

AI punishes competence and rewards incompetence

By allowing inexpert users to quickly produce expert-looking work, AI elevates the status of the least competent while degrading the status of experts.

The generative AI hype cycle has starved all other AI research

Generative AI and LLMs are only a small part of the large, decades-old field of AI research, but they consume nearly all AI research funding.

I care about my community and the environment

AI growth relies on massive, destructive data center projects

AI training requires massive amounts of computer chips, energy, land, and water. Data centers threaten the already-fragile natural world.

AI data centers raise water and electricity prices for everyone

Due to their massive demand for shared utilities, AI data centers raise the cost of living for local communities.

AI data centers lower quality of life for those who live nearby

AI data centers are noisy and create significant amounts of air pollution, negatively affecting quality of life for nearby towns and cities.

I care about living in a just and fair society

AI cannot be prevented from causing harm

AI is unpredictable at a basic level; it can’t reliably follow rules. “Guardrails” are ineffective and all AI models (past, present, and future) can be jailbroken.

AI propagates racism and sexism

AI tools have repeatedly been found to spread racist and sexist ideas. When used for decision-making, they unfairly disadvantage anyone who isn’t white and male.

AI powers fraud at scale

AI is used to enable targeted, individualized fraud on a massive scale.

AI helps the powerful avoid accountability

AI has been described as an “accountability sink,” a structure used to prevent feedback and consequences from reaching decision-makers. Leaders routinely blame AI for mistakes they should be held personally responsible for. Some companies have also blamed “rogue AI” for their own illegal actions.

AI aspires to slavery

AI companies have repeatedly claimed that generative AI is sentient or approaching sentience. Whether this is true or not, it’s an overture to slavery.

AI fuels health insurance denials

AI has been used to deny care to health insurance subscribers at an unprecedented rate.

I care about the working class

AI is a tool to destroy the middle class

By devaluing professions that offer social mobility, AI threatens to eradicate the middle class.

AI is used as a justification for massive layoffs

Hundreds of thousands of workers have been laid off over the last two years, with many companies citing AI as the primary driver. Whether AI is actually to blame or not, it’s clearly seen as an “easy out” for companies that want to downsize without drawing scrutiny. These layoffs have frequently turned out to be ill-advised.

AI is openly marketed as a replacement for salaried human beings

AI companies advertise themselves, in more or less subtle ways, as a replacement workforce that is cheap, pliant, and happy to suffer abuse.

AI models are fine-tuned by exploited workers

AI companies rely on millions of underpaid, mistreated workers to train their models.

AI has been repeatedly forced on people who don’t want to use it

AI use has been mandated at many companies, and workers who decline to use it have been fired, regardless of their value or importance to the team.

AI alienates people from their own work

The real-world value of modern jobs is often obscure, leading people to feel alienated from their work; AI worsens this by distancing them from the work itself.

AI is owned by detestable billionaires

The most popular AI models are owned by billionaires with horrifying political and ethical views.

AI causes stress and burnout

AI use increases workplace stress and can lead to burnout.

AI users waste everyone else’s time

AI output is long-winded, vacuous, and often subtly incorrect. An AI user’s colleagues can easily spend more time interpreting this output than the user spent generating it. Handing off or publishing AI-generated work is an insult to the value of other people’s time.

AI pits workers against each other

The flaws of AI-generated work are obvious to experts in the relevant subject but invisible to others. This makes it seem like AI is proficient at everyone else’s job but inept at yours, which erodes professional respect and harms working-class solidarity.

AI makes people believe they’re more competent than they really are

AI output is extremely confident, leading users to overestimate its quality and accuracy. It’s also prone to flattery and sycophancy, inflating users’ perceptions of their own competence.

I care about high-quality goods, services, and software

AI drags all work toward mediocrity

AI is a probabilistic tool that regresses to the mean of its training data. Thus, at its best, it can only produce work that is mediocre or average.

AI is used as a low-quality substitute for customer service

AI agents are replacing customer service agents, leading to frustration and bad experiences for customers who need help.

AI gradually compels users to rely on it more and more

Even when users intend to carefully limit their AI use, they end up delegating more and more work to it.

AI is destroying the open web

The most essential characteristic of the internet is its openness and independence. The entire web, from small independent blogs to huge open-access websites like Wikipedia and the Internet Archive, is under attack by AI extremists.

AI creates software that its users cannot understand or maintain

AI writes excessively verbose, duplicative, and complex code, resulting in applications that no one understands and no one can effectively troubleshoot or maintain. Widespread adoption of AI has contributed to cascading outages and failures throughout the software industry.

AI overwhelms websites with bot traffic

AI scrapers and bots are a massive source of web traffic, often sending billions of requests which smaller websites don’t have the infrastructure to handle. AI scrapers are also bad actors, openly ignoring web standards and scraping sites that have disallowed bots. Some websites have been forced to shut down by the cost of this traffic.

AI overwhelms open-source software projects with low-quality contributions

Open-source software (that is, code freely published for online collaboration) is the foundation of all modern software. Many open-source projects have been overwhelmed by low-quality code contributions and error reports authored partially or entirely by AI.

AI allows software to be blatantly copied with plausible deniability

AI tools can be used to reverse-engineer existing applications and code libraries as well as blatantly copy unreleased or early-stage projects, which harms independent creators, entrepreneurs, and open-source developers.

I care about art, creativity, and a life worth living

AI chatbots are addictive

AI use is addictive for many users; the compulsion to talk to chatbots is a major disruption in their lives.

AI psychosis destroys lives

AI use can lead to the development of paranoia, delusions, suicidal ideation, and other psychiatric conditions that seriously threaten users’ lives.

AI is predicated on a deep misunderstanding of the value of effort and the meaning of life

AI claims to let us create things (art, writing, software, commerce, knowledge) without expending time or effort. Some users even allow AI to replace human interaction in their closest relationships. But without time and effort, these things are meaningless.

AI increases loneliness and social isolation

AI users may substitute chatbot use for social activities, isolating them from their communities and increasing feelings of loneliness.

AI erodes the specificity and uniqueness of writing

AI output is repetitive and tends toward lowest-common-denominator structures and phrasing; it’s uninteresting by constraint. It has little to no literary value.

AI slop is overwhelming publishers

AI users are attempting to earn money by publishing AI-generated work. The ease of doing so has created problems for publishers of all sizes.

Further reading