The Risks of AI
Pull The Plug is not anti-AI. But we feel the risks of AI deserve attention and should factor into government decisions, ensured by the will of the people. People in Pull The Plug necessarily have different perspectives on AI. But we broadly agree that the below risks should inform policy and government decision-making.
## Introduction of bias
AI is capable of processing huge amounts of information, far more than any human, but that information is not neutral. AI learns from historical information created by biased humans and it inherits such biases, then amplifies them, leading to real-world discrimination. AI bias can feed into other models, consolidating and furthering bias.
We’ve seen that AI systems carry a deep-seated bias against [__women in the workplace__](https://www.gsb.stanford.edu/faculty-research/publications/age-gender-distortion-online-media-large-language-models), gender bias in [__social care systems__](https://link.springer.com/article/10.1186/s12911-025-03118-0), racial bias in [__AI image generators__](https://www.nature.com/articles/d41586-024-00674-9), racial bias in [__justice systems__](https://www.ajl.org/), and so on. AI models discriminate against just about every minority group and face very few consequences for presenting the bias.
## The rise of misinformation
Misinformation is nothing new. But AI pushes misinformation at a scale we’ve never seen. On the one hand, we’re seeing large language models (LLMs) incorporated on search engines, social media platforms, apps, and websites, all of which have the ability to hallucinate. The misinformation is often harmless, usually a fact that isn’t quite right or a made-up quote. But it can have devastating real-life consequences.
Reliance on chatbots has [__undermined justice__](https://www.theguardian.com/technology/2026/apr/22/ai-hallucinations-found-in-high-profile-wall-street-law-firm-filing) in legal cases. Google Overviews has offered [__dangerous and potentially life-threatening health advice__](https://www.sciencetimes.com/articles/50366/20240525/google-ai-ai-overviews-ai-generated-responses.htm). LLMs have promoted [__conspiracy theories__](https://www.cnbc.com/2024/05/24/google-criticized-as-ai-overview-makes-errors-like-saying-president-obama-is-muslim.html), published [__false academic sources__](https://www.nature.com/articles/s41598-023-41032-5), and undermined democratic processes. And the companies in charge of LLMs are not legally liable.
Then there’s the increasing problem of distinguishing the real from the unreal, pushed by deepfakes. We’ve seen deepfakes videos showing [__Vlodymyr Zelensky surrendering__](https://www.reuters.com/world/europe/deepfake-footage-purports-show-ukrainian-president-capitulating-2022-03-16/) and [__Donald Trump being arrested__](https://www.bbc.co.uk/news/world-us-canada-65069316). Fabricated voice recordings seemed to show a [__Slovakian politician__](https://edition.cnn.com/2024/02/01/politics/election-deepfake-threats-invs) rigging an election and [__Joe Biden urging voters not to vote__](https://www.bbc.co.uk/news/world-us-canada-68064247). All of these undermine political systems and contribute to rising misinformation.
## Copyright theft
The list of lawsuits against AI companies could fill a book. Judi Picoult, John Grisham, and George R. R. Martin – who once called generative AI “the world’s most expensive plagiarism machine” – [__have taken OpenAI__](https://www.bbc.co.uk/news/technology-66866577) to court. Major publishers, including Penguin Random House and HarperCollins, [__recently sued__](https://www.reuters.com/legal/litigation/publishers-sue-shadow-library-allegedly-powering-ai-chatbots-2026-03-06/) the shadow library Anna’s Archive, claiming it distributed pirated books that major AI companies used for training. In 2026, roughly 10,000 writers, including Kazuo Ishiguro and Malorie Blackman, published [__Don’t Steal This Book__](https://www.dontstealthisbook.com/), an ‘empty’ book that showed only a list of their names, a protest against AI firms using work without permission.
LLMs have been trained on huge amounts of data, much of which relies on copyrighted books, articles, images, music, and videos. Original authors and creators have not been asked for permission and rarely receive credit or compensation, but Big Tech companies profit from their work. And generated outputs undermine the jobs of creators, using shallow imitations of their work to remove future opportunities. Copyright theft poses a threat to flourishing creative activity.
## Economic inequality
Economic policies should reduce uneven distributions of wealth. A common criticism of AI is that companies can reduce dependence on a human workforce, as above, and increase the wealth gap. The economic rewards of AI are concentrated among tech companies and people working on the systems. At the same time, workers face pressure from automation, creating stagnation for middle- and lower-class groups.
We’re seeing the early impacts of AI and inequality. [__Research from MIT__](https://shapingwork.mit.edu/research/the-simple-macroeconomics-of-ai/) suggests that AI advances will widen the gap between capital and labour income, already a key source of inequality. Analysis from the IMF and [__from CEPR__](https://cepr.org/voxeu/columns/ai-and-distribution-income-between-capital-and-labour) echoes the point that AI shifts returns towards capital, which leads to greater concentrations of power.
The true impact of AI on the economy is not yet known. But it is likely that, without intervention, without access to reskilling and fairer tax systems, AI will deepen social and economic divides. We should aim for democratic accountability for the economy.
## Environmental decline
AI systems are driven by data. Every time anyone uses an AI model, the data passes through servers housed in data centres. The increasing use and mis-use of AI has boosted demand for data centres, with the UK aiming to build more than 100 new ones on top of the 500 already built. Data centres demand huge amounts of electricity and water. Even a small data centre can use enough electricity to [__power 1,000 houses__](https://eng.ox.ac.uk/case-studies/the-true-cost-of-water-guzzling-data-centres/) and [__25.5 million litres of water annually__](https://www.arup.com/insights/how-can-we-cut-water-consumption-in-data-centres/). And, by 2030, projections suggest that data centres will use more electricity than the entirety of Japan.
The ever-growing use of AI is unsustainable, in every sense. AI adoption and increasing demands for computation harms the environment, particularly because it is largely powered by fossil fuels. Big Tech companies have proved particularly opaque around carbon emissions. Some have given the wrong information, [__understating emissions__](https://www.theguardian.com/technology/2026/may/09/google-developers-significantly-misstate-carbon-emissions-of-proposed-uk-datacentres), and others refuse to [__disclose environmental costs__](https://news.mit.edu/2025/explained-generative-ai-environmental-impact-0117). We should have transparency around the environmental cost of AI models.
## Pressure on public services
Data centres are necessary for the implementation of AI. Governments across the world have prioritised the building of data centres, often at the expense of necessary public services. The British government classifies data centres as [__Critical National Infrastructure__](https://questions-statements.parliament.uk/written-statements/detail/2024-09-12/hcws89), which means they often jump the queue ahead of houses and hospitals. AI and data centres contribute to the housing crisis.
AI puts pressure on housing and hospitals, and pushes up costs, with energy-hungry data centres increasing our electricity and water bills. Data centres force expensive grid upgrades and demand improvements to water systems, both of which come out of your energy bills. The public did not consent to AI taking priority over hospitals and houses, nor did we consent to cost of living increases, and we should have a say.
## The loss of control
AI has been used in our [__courts__](https://cdn.prod.website-files.com/67becde70dae19a9e5ea2bc3/689e037a14d556416d51b37b_AI-in-our-Justice-System-final-report%20(1).pdf), our [__army__](https://ukstopkillerrobots.org.uk/2025/09/04/uk-crossing-the-line-as-it-implements-use-of-ai-for-lethal-targeting-under-project-asgard/), our [__police__](https://www.apccs.police.uk/apcc-statement-on-facial-recognition-evaluation-report/), and across [__central__](https://bigbrotherwatch.org.uk/press-coverage/big-issue-unreliable-ai-usage-by-dwp-risks-vulnerable-people-being-treated-as-guinea-pigs/) and local government. But there are very real concerns, voiced by [__Nobel Prize winners__](https://www.theguardian.com/technology/2024/dec/27/godfather-of-ai-raises-odds-of-the-technology-wiping-out-humanity-over-next-30-years) and [__AI experts__](https://www.cam.ac.uk/stories/malicious-ai-report), that we could lose control of the tech to devastating effect. Already we’re seeing powerful AI systems that are difficult to predict, interpret, or understand, an issue known as the [__‘Black Box’ problem__](https://towardsdatascience.com/the-black-box-problem-why-ai-generated-code-stops-being-maintainable/). Future consequences are simply unknown.
We need to regain control of AI. We need the government to put preventative measures in place to stop the development of uncontrollable AI. It is absurd that, in the UK, AI is less regulated than a sandwich. For the sake of humans, and future generations, the people need to decide how to regulate AI ethically and responsibly.
## Job displacement
The extent of job displacement is hard to quantify, but we’re already seeing AI shape the job market. [__Research from the US__](https://www.goldmansachs.com/insights/articles/how-will-ai-affect-the-us-labor-market) suggests that AI could replace 7% of workers over the next decade. A major [__UK study__](https://www.theguardian.com/technology/2024/mar/27/ai-apocalypse-could-take-away-almost-8m-jobs-in-uk-says-report) found that Britain could face nearly 8 million job losses. [__Analysis from Brookings__](https://www.brookings.edu/articles/measuring-us-workers-capacity-to-adapt-to-ai-driven-job-displacement/) suggests that young people and women will be disproportionately impacted. And a [__recent Gartner survey__](https://www.gartner.com/en/newsroom/press-releases/2026-05-05-gartner-says-autonomous-business-and-artificial-intelligence-layoffs-may-create-budget-room-but-do-not-deliver-returns) demonstrates that companies will use the tech as an excuse to downsize, which we’ve already seen, particularly in the Big Tech space that fails to show evidence of productivity gains.
Others argue that AI is just ‘creative destruction’, the process of industrial mutation, destroying old structures and creating new resources. Job losses will lead to the creation of new roles. But even in that best-case scenario, the transition will be uneven unless we take action. Some workers might find new roles, while others face barriers, such as lack of access to training or digital exclusion. AI job displacement is not a tech issue, but a social issue and one over which the public should have a say.
## How you can get involved
There are plenty of ways to get involved. We’d love for you to join and help us Pull The Plug on Big Tech. You can get started by…
* Signing our petition demanding AI regulation
* Joining the [__Pull The Plug campaign__](https://pulltheplug.uk/sign-up/)
* Joining our call for Citizens’ Assemblies