AI is now so widely contested we can barely keep up. That’s good! See below for what we managed to track for the period May to July 2026:

One we missed when it was first shared with us earlier this year – sorry it’s late! –  AI industry insiders launch site to poison the data that feeds them

The movement against data centres is growing, not only in the US (although it’s also very large in the US): Data Center Watch reports that local activist groups in the US have doubled to 833 nationwide and have blocked or delayed at least 75 distinct AI data centers projects worth a combined $130 billion in Q1 2026. These figures are roughly the same – in the first quarter of 2026 – to the whole of 2025. As this survey shows, much of this opposition to data centres reflects an underlying opposition to AI more generally. The Guardian published this article on the anti-datacentre movement. Good analysis here from Brian Merchant on the dramatic rise in opposition to data centers in the US, especially amongst poorer communities. The AI Now Institute continues its series of workshops advising activists how to challenge data centres, including on protecting water resources. Social Change Lab published this excellent report documenting how campaigns against data centres form part of the climate movement. Erin Brockovich has built a map of AI datacentres in the US , marked with concerned raised by residents. Heatmap reports at least 20 data centre projects cancelled due to local resident opposition in the first quarter of 2026, accounting for more than $41.7 billion in investment and represented at least 3.5 gigawatts of electricity demand. This includes Project Jarvis, which was set to use up to 1 gigawatt of electricity but lost a vote in the planning board and then withdrew the application altogether following Ron DeSantis’s proposals for state-level regulation. One survey found Americans would rather have a nuclear plant in their backyard than a datacenter. To get an idea how angry data centres are making people, especially in the US  – watch this. Overall the scale of opposition to data centres is causing considerable delays in construction. Corporations have tried to blame anti-data centre sentiment on China (lol).

On specific campaigns against data centres in the US, see: this campaign in Ypsilanti Township, Michigan, against a proposed data centre.  Reno City Council voted to keep a temporary data center moratorium in place until Aug. 31, 2027 – a move welcomed by campaign group the Water Over Data Alliance.  Warren Township, NJ, (Republican controlled) voted to ban data centres. Imperial County, California, adopted a 45 day moratorium on data centres. Dane County, Wisconsin, voted for a moratorium on data centres for 18 months. More than 300 Pittsburg residents showed up to a City Council meeting in support of banning a data center development. Seattle announced a 1-year moratorium on data centres. Residents showed up to protest a planning meeting for a data centre in Allen Park, Michigan – with the application rejected. Residents attended a town hall at Kenilworth, NJ, to oppose a data center. Monterey Park became the first city to vote to ban data centers in a public vote – with 86% of voters supporting the ban. An interview here with a local activist who helped to block a data centre in Fayetteville, Georgia. Residents oppose data centre in Hillsboro City, Oregon. Residents are suing Birmingham, Alabama, and AI company Nebius Inc. over a planned data centre. A 2-year legal battle over a planned data centre in Prince William County, Virginia, resutled in the successful blocking of the scheme after the company, QTS, opted not to pursue an appeal after its planning permission was voided in the Virginia Circuit Court. Resident opposition to AI surveillance in Troy, New York, led to the mayor declaring a state of emergency to override the objections of both the public and city officials. Charlotte City Council voted to establish a moratorium on data centers for 150 days (community response to the decision here). Republican member of Congress, Nancy Mace, called for a one-year moratorium on new data center construction in South Carolina. Cullman City Council adopted a temporary moratorium on data centers. Saline Township, Michigan, has seen construction start on a 250 acre AI data centre being built for Oracle with enough power for 1 million homes. This is prompting considerable local resident opposition, with NDAs preventing residents from knowing the full details – NYT has some detailed reporting (including on the electoral fallout) here. The Township Treasurer later resigned due to the scale of public hostility (including death threats). The QTS data center,in Fayetteville, Georgia, is being investigated after it is alleged it drained 30 million gallons of water during a drought – see the angry residents here. Andover Township banned data centres (see the response of residents here) after an earlier town hall meeting resulted in an angry resident being attacked by the police. Residents at Red Oak, Texas, protested plans for a 800-acre data center, packing the City Hall meeting. Residents of Van Buren Township expressed outrage about a planned hyperscale data centre. New York has become the first US state to pause new data centres for one year. St. George City, Missouri, voted for a permanent ban on data centers –  also prompting St. George County to also adopt a 6 month ban – good account here on how the local campaign was organised. The website Moratorium Nation is tracking the moratoriums currently in place in the US.

Outside of the US: A proposed data center in Katoomba, New South Wales, Australia, was met with strong opposition from residents and local authorities and subsequently withdrawn. Protests in Vizag, India, against a planned data centre. LSE published a report highlighting a growth in legal cases against data centres due to their environmental impact. 13,000 residents in Inzai City’s Chiba New Town Chuo, near Tokyo, Japan, signed a petition against a new planned data centre.  Dutch activists Geef Tegengas occupied and blocked access to a major Microsoft data centre construction site in Amsterdam, escalating a growing campaign against the tech giant’s role in Israel’s military and surveillance infrastructure. Manitoba Premier Wab Kinew announced that Manitoba would reject a large AI data centre proposed for farmland south of Winnipeg. Hamilton City, Ontario, saw a large protest against a planned data center – prompting the planning tribunal to reject their application. There is an ongoing campaign and demonstrations to oppose a data centre planned for Brick Lane, London – see James Meadway interviewed on the campaign here, and a load of photos from a recent demonstration, organised by Pull the Plug, here, and some speeches from the protest. There is a campaign page here. The UK Government’s AI growth zone in Lanarkshire, Scotland, came under criticism with the countryside charity Action to Protect Rural Scotland (APRS) raising concerns over the expansion of data centres on the site. Edinburgh City Council moved to consider a data centre moratorium as they experience pressure from the Scottish Government to approve the planning applications – which is part of a proposal for 5 data centres across Scotland. More here on the campaign in Scotland with APRS now saying there are 14 projects in the pipeline. In the end, Edinburgh City Council decided a moratorium was not legally possible, so it opted instead to vote to call on the Scottish Government to bring in a moratorium on new data centres. Residents in Cardiff, Wales, are opposing a planned data centre – although the city councillors are trying their best to ignore the residents. A campaign has been launched against plans for a massive data centre in Great Torrington, North Devon. Protesters from Save the Lammermuirs – Stop the Data Centre (SLSDC) are asking for a rethink of the “monster” Southside Data Centre in the hills near Duns, Scotland, between the villages of Longformacus and Westruther (more on this story here). Also in Scotland, hundreds of people packed into a Fife village hall to oppose proposals for a massive AI data centre. Residents in Reading, UK, are opposing a planned new data centre. The UK Government is pressuring local councils to grant planning permission to data centres despite resident opposition, for instance in Epping Forest, and Kent. 5000 people signed a petition against a planned data centre in Norwich, UK. Vancouver saw two well attended demonstrations against data centres in the space of a month – here (and some interviews here with the protesters) and here. 

Public opinion is generally against AI, especially in the US and amongst young people: Hasan Minhaj has a good discussion here on the Gen Z backlash to Big Tech. A survey of GenZs found only 18% were hopeful about the technology – more discussion of that survey here.  Another survey showing similar results here. And another survey showing widespread opposition to AI. California primary elections saw tech-backed candidates do poorly. 

The University graduation period witnessed a spate of booing by students when speakers mentioned AI: first, at University of Central Florida, but then at a series of universities, including Marquette University (in opposition to the speech given by Adobe Inc.’s global head of AI and agentic systems, Chris Duffey), University of Arizona (against Google Chief Executive Eric Schmidt), Glendale Community College (against the College’s use of AI to read out student names, which it messed up), Middle Tennessee State University (against Scott Borchetta, CEO of record label Big Machine, after he told disgruntled graduates to “Deal with it”). In contrast, Ronny Chieng told Harvard Graduates “F*** AI” and that it is ‘stupid’, to a lot of cheers. Students at Stanford walked out in protest at Google CEO Sundar Pichai’s speech, citing Google’s role in the genocide in Gaza. 

Some attempts to address the impact of AI on higher education: The Leiden Declaration saw mathematicians raise concerns over the impact of AI. Pull the Plug are organising for an AI-free spaces in universities.

AI firms continue to face legal issues: Publishers of nearly 400 newspapers are suing Open AI and Microsoft for scraping their content in training their AI models. Florida has moved to sue OpenAI and Sam Altman over the unsafe nature of ChatGPT and the fact that it is hurting children. A court in eastern China ruled in favor of a senior tech worker whose company replaced him with artificial intelligence, apparently creating a much-discussed precedent that would protect workers from dismissal due to AI in China. In the Musk vs. Open AI trial, ex-OpenAI exec Sutskever said he spent a year gathering proof of Sam Altman’s dishonesty, and Mira Murati, OpenAI’s former CTO, also testified that she couldn’t trust Altman’s words. Apple agreed to settle a lawsuit after it was accused of misleading customers about the capability of its AI features – more detail here on the $250m settlement. Five publishing houses and author Scott Turow sued Meta and CEO Mark Zuckerberg, over its use of copyrighted works to train its AI language system Llama. Jess Asato MP is suing xAI over Grok’s nudification function.  OpenAI is facing another lawsuit after ChatGPT told a 19-year-old, Sam Nelson, to take a lethal mix of Kratom and Xanax. Texas launched an investigation into Meta’s AI glasses over privacy and data concerns.

Palantir is becoming increasingly unpopular: Immigrant rights lawyers have sued Palantir over its ELITE Tool which is used by ICE to track immigrants in the US. Sheffield City Council voted to call for local NHS trusts to reject the use of Palantir for data processing. In the UK opposition continues to grow to Palantir’s NHS Federated Data System. The Parliament’s Technology Committee recommended that Palantir should not have a significant role in government data systems. Pull the Plug staged a lively demonstration outside the NHS’s largest conference of the year, in Manchester, protesting Palantir’s role in the NHS. Palantir’s case wasn’t helped by the recent discovery that its data is full of errors. France opted to end any use of Palantir data tools for security reasons. Spain have also ended the use of Palantir. Pull the Plug pulled off a nice stunt opposing Palantir with the delivery of 60 boxes of “data” to the company. 

A range of anti-AI protests are being regularly staged: An anti-ICE protester was asked to leave the Spain vs. Saudi Arabia match. An anti-AI march was staged at Granville Island, Vancouver. Meta laid off 1400 workers and then Zuckerberg went on a nice trip on his yacht – here is the protest outside the yacht. SEIU members in New York protested outside Palantir HQ in opposition to the companies role in public surveillance. San Fransisco saw a march from the HQ of Open AI to Anthropic protesting AI under the slogan, “Stop the AI Race”.

The Green Party in the UK has launched a ‘Say No to Waymo’ campaign in Harlesden, London, where the self-driving cars are being piloted – see interviews with the campaign organisers here – the campaign is in partnership with the App Drivers and Couriers Union and Pull the Plug.

Workers continue to organise against AI: NYT employees are organising against the use of AI to monitor their performance. The UPTE union has created the largest ever tech workers union in the United States at the University of California – as they say in their press release their fight “increasingly centers on artificial intelligence. UPTE members will gain the formal right to bargain collectively over working conditions, including how UC deploys AI tools. The vote arrives as private tech employers have cut thousands of jobs nationally, blaming AI automation to justify workforce reductions”. – see a write-up here from Brian Merchant. UK staff of Google’s AI research lab, Google DeepMind, voted to unionise after concerns were raised about the move to militarize AI models. Meta employees distributed flyers in their workplace opposing the introduction of new mouse-tracking surveillance by the company – before eventually being allowed to opt out of the surveillance (but only for 30 minutes). The threat of strike action in South Korea by unionised workers at Samsung was averted as a last minute deal prevented strike action and was brokered by Labour Minister Kim Young-hoon – with the workers seeking a share of the massive profits reaped by Samsung due to the massive demand for semiconductors due to the AI boom. Discussion here by Rest of World (and what it might augur for workers generally) here. The deal was widely reported to have included an average bonus of $400,000 for each employee – although this was concentrated on the semiconductor workers, causing division with other workers in Samsung who got in some case much less, resulting in an exit from the union – see Bloomberg’s detailed write-up here.

Workers are also taking up less visible acts of dissent: Google employees leaked the memes they are sharing on how AI sucks. Amazon staff were reported to be purposefully over-using AI tokens in an attempt to appear more active than they were.

The flawed and counter-productive effects of AI continue to be highlighted: Research shows AI has negative psychological consequences for workers – undermining productivity gains. Likewise the constant need to correct AI errors negates productivity gains according to the Work AI Institute. AI is also slowing the job recruitment process. It also degrades the abilities of physicians and software engineers. Nature retracted a paper that claimed AI had a positive impact on student learning. KPMG published a report on AI that was itself full of AI hallucinations.  Rape convictions were called into question in Derbyshire, UK after it transpired that AI had been used in crafting the paperwork. Research from Cornell University found it was extremely easy to use Reddit to manipulate AI agents. The use of AI to draft lawsuits is throwing up a range of problems, including halucinations and lack of clarity over issues of liability and confidentiality. See a video of judges ripping into lawyers who used AI to create fake legal arguments here. A Brazilian court fined two lawyers for using hidden text in a document in an attempt to influence the court’s AI tools. Ford opted to re-hire expert engineers after AI failed to deliver accurate results.  Waymo was forced to recall 3,800 robotaxis after one drove into a flood. Software developers reported that AI is rotting their brain.  Rest of World have a story covering five times AI embarrased governments as a result of its errors. AI-driven healthcare provision in Kenya is driving up costs for the poor – the exact opposite of what it promised to do. Doctors’ AI note takers routinely make mistakes.  Meta has withdrawn its AI Muse Image function after it was used (entirely predictably) as a nudification app. A recent study found you can persuade AI models to accept falsehoods as truth. Hackers used Meta AI to access instagram accounts. Attempts to turn the AI companion app, Character.ai, into a greater revenue-generating app resulted in enshittification with widespread outcry from users. Myer’s Bagels in Burlington, Vermont, received considerable backlash when it used AI for its advertising.  Waymo was forced to apologise after its car repeatedly (and noisly) got stuck in a cul-de-sac in London. Scores of Waymos got stuck in a single cul-de-sac in Atlanta suburb. Cisco used AI to write security incident reports – routinely including errors. Stanford study confirms what we already knew – AI-led job recruitment is racist. Lawyers from UK law firm Pinsent Masons were found to have made false submissions to a judge based on AI.

Questions continue to be raised about whether AI really can raise productivity in the way that it was promised it might. Having once sought to push employees to engage in ‘tokenmaxxing’ companies have started moving in the opposite direction – towards ‘tokenminning’! due to the costs of tokens – and now doing their best to slam on the brakes! (including by making AI talk like cavepeople). For instance, Uber’s COO went on the record to say that the costs of tokenmaxxing were difficult to justify – and Amazon scrapped its AI leaderboard which had tried to encourage employees to engage in tokenmaxxing.

There have been a range of both campaigning and academic events critical of AI: New School organised a Luddite Conference on Participatory Futures.  The Campaign against Climate Change and the Greater Manchester Climate Justice organised a conference, This Is Not Inevitable: Standing up to AI and Big Tech.   CrAIc at Exeter hosted an inspiring workshop on what it means to do critical AI studies. Contentious Politics of AI Network co-founder, Iran Mansouri, took part in an online seminar, Colonialism in the Digital Age, alongside Dan McQuillan and Patrick Brodie.  Brian Merchant has an excellent report on the recent Summer of Ludd, a series of protests, performances, talks, and meetups in New York City organized around the idea that we need to get people into the streets and break Silicon Valley’s grip over our lives.

StopGenAI is putting on a series of webinars on the topic: Recovering after Gen AI.

Some new regulations and other initiatives to tame AI: South Korea adopted new regulations for data protection relating to AI, as did Vietnam and China. Japan is considering a range of AI safety legislation. The  EU agreed to ban nudification apps. In early May 2026 it was mooted that Trump was considering introducing AI safety regulations. This led to an executive order to regulate AI nearly being signed in late May before finally being with withdrawn at the last minute seemingly due to industry lobbying – WaPo has a good overview of the powers that blocked it here (incl. Musk, Zuckerberg and David Sacks), and a similar discussion from Politico. He then signed a much lighter version of the executive order in early June (another report here in NYT) (IBM approves). California Governor Gavin Newsome signed a new executive order intended to distribute AI benefits to workers and small businesses.  The UK Competition and Markets Authority (CMA) has announced that publishers can choose not to appear in the AI Overviews of Google search results in the UK. The Oscars banned AI from winning acting and writing awards. A new US Guard Act is being proposed to ban minors from using companion chatbots, and a similar Chatbot Act is also being considered. The UK government continues to propose a ban on social media for under-16s. The EU ratified the Council of Europe Framework Convention on Artificial Intelligence. Illionois passed one of the strongest AI safety bills. ArXiv announces it will ban authors for a year if they are found to have used unverified AI for their papers. MPs in Britain are trying to introduce a ‘kill switch’ into legislation which would allow data centres and aI models to be shut down in the event of an emergency. The  UN says AI might heighten inequality (can you believe it!?) – see the Secretary General’s speech here. The G7 issued a statement declaring the need for safe AI – although the US continues to block attempst at coordinated regulation. 

Pope Leo XIV issued his encyclical calling for safer use of AI and calls for AI to be disarmed – see an overview from Washington Post here – and a discussion of the relationship between the Vatican and Anthropic that is part of the story – and an insider account from a member of the panel that informed the encyclical – for Brian Merchant this represented Anthropic successfully duping the Pope.

New anti-AI organising tools: StopGenAI has a guide for googling now that google is infected with AI, and a long list of easy things you can do fight GenAI. DAIR launched its Luddite Lab Worker Resource Hub. An article in The Conversation describes how citizens can resist AI by ‘data poisoning’.  Brian Merchant has some advice on how to de-google your life. The AI Resist List was launched by Karen Hao and others. Waging Nonviolence published a piece outlining 10 reasons to resist AI.

Some noteworthy political developments: A Congressional Democratic Party primary in Manhattan saw a standoff between two candidates, one (Micah Lasher) backed by pro-AI Super PAC (Leading the Future) and Mike Bloomberg, who beat AI Safety candidate, Alex Bores – in what was widely seen as a test of the AI industry’s ability to influence elections through big spending (which they did) (covered here). See here for an overview of the broader battle of the Super Pacs.  Good summary here of some of the Democratic Party’s more sceptical positions on AI, especially around Sanders and Warren. Xavier Becerra, running for the Democratic Party candidate for the upcoming California Governor election, proposed an 11-point plan to regulate AI and introduce guardrails for workers and kids. 

Concern continues to be raised about the potential for AI to produce a bubble: The BIS has raised the alarm about potential over-investment (and the US Treasury is privately also concerned -as are seemingly the holders of AI long-term debt).

In terms of alternative forms of AI, Scholz and Tortorici argue for cooperatives as an alternative to a corporate model of AI.

And two interesting proposals for sharing the wealth accumulated by AI: South Korea policymaker, Kim Yong-beom, chief of policy in the administration of President Lee Jae Myung, floated the idea of what he called a “people’s dividend”, whereby members of the public should be paid a share of chipmakers’ massive profits (more coverage here). Bernie Sanders is calling for the partial nationalisation of AI companies (which seems to have the support of two-thirds of US citizens).

The resistance to AI continues!

Workers against AI. Meta is relying on a complex performance indicator method to force employees to use more AI. Health workers at Kaiser Permanente went on strike over being instructed to use AI. A range of tech worker organisations opposed War Secretary Hegseth’s attempt to force Anthropic to acquiesce to Pentagon’s demands for full cooperation with any and all forms of repression. NHS staff are boycotting Palantir’s Federated Data Platform – although the NHS data chief remains undeterred. Nearly 600 staff at the University of Edinburgh signed an open letter calling for the end to a contract with OpenAI. Translators in Japan spoke out about the negative impact that AI is having upon their work. Voice actors who dub films into native languages are mobilising to avoid AI replacing their jobs – including in Brazil, South Korea, and China – and with Mexico also banning the technology for dubbing. Rafael Grohmann has built a tracker to map such initiatives, listing more than 100 movements by creative workers in about 25 countries. Authors continue to pursue Anthropic for a share of the compensation agreed with the firm for its use of their work in training its model with 120,000 authors and other copyright holders signing up for a share of the compensation (covering 91% of the books named in the settlement). A recent survey found 54 percent of workers avoiding their company’s in-house AI tools in order to complete tasks themselves. Chinese tech workers are being asked to train their AI doubled to perform their work flow, prompting innovative methods of sabotage. Palantir’s employees are expressing doubt over the firms products, especially after its central role in the ICE operations. GenZ workers are especially negative about AI – in a recent survey, employed Gen Zers were more than three times as likely (48%) to say the risks of AI in the workforce outweighed the potential benefits than to say the reverse (15%). About 30,000 union members rallied outside Samsung’s semiconductor plant in South Korea, calling for a greater share of AI profits. 560 employees of Google wrote to the CEO, Sundar Pichai, calling for the company to block AI military use – also covered here.     

Opposition to AI’s misogyny: The Dutch courts banned Grok from distributing non-consensual AI-generated naked images of women. Narwhal Labs’ ad for ‘AI employee’ in Bristol airport sparked outrage and complaints for its mysognistic content, eventually being removed. 

Faith groups against AI: wrote to the Armed Services Committee of both the US House of Representatives and Senate calling for boundaries on how the Pentagon uses AI for military purposes.

Voters against AI. David Sacks-backed Innovation Council Action announced plans to spend $100m on campaigning for AI in the face of voter opposition to the new technology. Concern over AI, this piece argues, is a position that unites AI safety billionaires and the left. Only 15 percent of Americans would trust AI as their boss, and 65 percent of Americans oppose building an AI datacenter in their community. Public opinion tends to be sceptical about the potential for AI to have a positive effect in a range of areas, including elections, the quality of news, the criminal justice system, and the education system. More discussion of these negative trends in public opinion here. Voters in Festus, Missouri voted out its councillors who had backed a new data centre in the town. Over 200,000 people in Britain signed a petition calling for an end to the government’s contract with Palantir. Protesters gathered outside Anthropic’s headquarters in San Francisco to call on major AI companies to temporarily halt the development of AI systems.

Politicians on AI: Given the growing groundswell of opinion against AI, this is translating into some opposition at the political level in the US. The US Democratic Party has however seen its opposition somewhat (but not entirely) stymied by pro-AI PACs funding campaigns against politicians who seek AI regulation – AOC has promised not to take any funding from the AI industry.  Democratic House candidate in New York, Alex Bores, is proposing an AI dividend that would charge AI firms for their success so that it is shared with US voters. Trump’s Executive Order seeking to prevent states from regulating AI has so far failed to miss its own deadlines.

A number of US states have moved to regulate or restrict AI. California – Governor Newsom signed an executive order on Monday that gives California four months to develop AI policies that prioritize public safety. Plans for a new AI school in Manhattan were cancelled at the last minute due to growing unpopularity and concerns about the impact of AI on education. Colarado has adopted an anti-discrimination bill to regulate AI – prompting xAI to sue them, along with the Justice Department. Florida Attorney General announced the start of a criminal investigation into OpenAI and ChatGPT over its potential role in the Florida State University shootings last year. Illinois is proposing an AI law that would shield AI firms from liability – prompting opposition from Anthropic. More on the liability question here.

Other moves for further regulation outside the US: The Indian government proposed legislation that would require platforms to display a clear label on AI-generated content for the entire duration of the content. The Dutch government proposed an AI ​​Regulation Implementation Act. The British government launched consultation over regulation for automated decision-making/profiling. EU staff were banned from using AI-generated content in official communications. The UK parliament’s science, innovation and technology select committee tried to find a way out of the Federated Data Platform, much to Palantir’s chagrin. 

Users against AI: 125,000 Bluesky users blocked its new AI assistant, Attie, within days of its launch. 

The Pope called for more regulations on AI. 

Local residents, farmers, and indigenous peoples against AI data centres: Loughton Town council moved to stall the construction of a new mega-data centre in Essex, UK. In the US, Wall Street has noticed that data centres are unpopular and being stalled by local opposition.  This situation is rapidly changing. By early April, 11 states in the US had proposed some legislation to restrict or ban data center development since late 2025. This followed Bernie Sanders and Alexandria Ocasio-Cortez’s call for a moratorium on Data Centres. Maine became the first state to ban data centres – part of a growing trends of states considering the move. A later estimate cited twenty-seven states advancing data center legislation that requires developers to cover data center energy costs and report usage. Dutch farmers are opposing the construction of data centres, with opposition consolidating in a new popular party, the Famer Citizen Movement. India is also seeing opposition from farmers who oppose the government’s facilitation of sale of land to data centres. One man in Oklahoma was arrested for speaking against data centres for several seconds too long at Claremont City Council meeting. Prince William County ended plans for a very large data center project near Manassas National Battlefield Park – this followed polls showing a more than doubling of public opposition to data centres in Virginia, the US state with the largest number of data centres, over the past two years. The NAACP is suing xAI over its data centre pollution in South Memphis. As the FT reports, this local opposition is slowing construction of data centres in the US.  Montery Park residents successfully prevented a data centre in their locality – some reflections here on lessons from that campaign, and some more discussion here. The indigenous Anacé people in Brazil is opposing a planned new data centre for Tik Tok. The Ypsilanti Township in Michigan voted for a 365 day moratorium on the delivery of water to hyperscale data centers. Residents are campaigning against a plan to build six sprawling data center campuses in Archbald, Pennsylvania. Honor the Earth, an Indigenous-led environmental justice organization is tracking over 100 proposed data center projects on tribal and rural lands – further coverage here. The Seminole Nation of Oklahoma became one of the first Native American Tribes to enact a complete moratorium on data center development within its jurisdiction.

The new North Star Data Centre Policy Toolkit presents excellent advice for local campaign groups seeking to oppose data centres. This is accompanied by a series of talks on youtube presenting the tool kit and campaign advice. On one of the campaigns featured – sign the petition to oppose Data Center at Former Landover Mall Site.

Politicians on data centres: The Republicans are split over how to respond to the public concern over data centres – between seeking to regulate in order to avoid higher energy costs, and the more generally pro-AI stance of the Trump administration. Democrat House Minority Leader Hakeem Jeffries indicated that the Democrats would move to shield consumers from the energy cost rises associated with datacentres. 

Parents called again for the US Congress to put more safeguards for children using AI. Families moved to sue OpenAI over Canadian mass shooter’s use of ChatGPT.

Investors are getting anxious about the amount of debt being taken on to fund data centre construction, pushing for a higher yield.

Musicians opposing AI – listen to Reject the Machine; and Taylor Swift filed for copyright for trademarks, to avoid AI replicas

Campaign groups opposing AI: The Parents and Kids Safe AI Coalition was found to be heavily supported by OpenAI, prompting many activists in the group to leave. There’s a lengthy discussion of campaign group ControlAI in the Washington Post here. Corporate and Europe Observatory and AlgorithmWatch conducted research for an expose published by Investigate Europe that highlights how Microsoft ghostwrote EU policy that keeps data centres’ energy use secret. A new Purge Palantir group has formed. The Cables of Resistance conference took place in Berlin and was well attended and well received in the press – see here. Foxglove, a tech-focused campaign group helped uncover the UK government’s under-estimation of AI data centre emissions. Resisting Big Tech, a full-day conference organised by Global Justice Now, took place in London – watch the plenary here.

Journalists on AI: Ronan Farrow wrote an expose on Sam Altman’s proclivity for lying. 

Elon Musk, owner of Grok, wants a non-profit OpenAI: as we saw in the high profile legal case, some coverage here, and discussion of the awkward comments revealed as part of the legal case.

Concerns continue to be raised over the surveillance implications of AI: report here on a plan for AI-driven surveillance in Toronto’s wealthy Rosedale neighbourhood which is prompting concerns. 

Sam Altman’s house was attacked by a man with a Molotov cocktail, apparently by someone who is fearful of the effect of AI – more details here and here. At the same weekend an Indianapolis councilman had shots fired at his house and a ‘no data centers’ note left on his doorstep, after the city’s Metropolitan Development Commission approved the rezoning of the land to make way for a new data centre. This violent turn is discussed more by Brian Merchant here.

And anyway AI doesn’t work (at best): Robotaxis froze unexpectedly in Wuhan, and in a Wizard of Oz moment Tesla were forced to admit its robotaxis are sometimes diven by remote humans.  A recent study into the use of AI for health advice found people who used AI chatbots were less likely to identify the correct health condition than those who investigated the problem without AI. Another study found thatAI chatbots misdiagnose in over 80% of early medical cases – more coverage here and here. Another study found that ChatGPT makes you less clever – and another found it prevents people from thinking critically. Another found that AI agents perform at roughly 50% of the level of reliability as human specialists with a PhD. Shy Girl by Mia Ballard was pulled by Hachette Book Group following accusations that the book was AI-generated. And another study found AI creates ‘brain fry’. Google’s AI Overviews provides tens of millions of erroneous answers every hour. One of Taiwan’s top newspapers published a political story fabricated by a chatbot. Facial recognition led to a woman being wrongly jailed for 6 months in Maryland. A chatbot was used to generate fictitious AI-generated complaints that led to the shut down of Heaven night club in London. Anthropic’s new Mythos AI model sparked global concerns about the risks that it poses. And see here for a discussion of the rise of fake cases entering the legal system due to AI.

Some coverage of social-oriented alternative forms of AI, covered in Rest of News here, might be of interest.

For more on resistance to AI: follow the Pivot to AI youtube series; NYT published a long read on opposition to AI.

Maybe relevant: China banned Meta from buying Chinese AI company, Manus – discussed in more detail in FT here.

The resistance to AI continues!

Campaign updates:

• April 25, 2026 – Resisting Big Tech Empires – London South Bank University – details here
• London saw the biggest anti-AI protest ever. Reports here and here.
• Pull the Plug recently postered the OpenAI office in London to protest their role in AI-powered warfare.
• NYC saw mass vandalizing of adverts for the AI wearable Friend.
• Humans First was launched – a new AI-safety social movement
• The QuitGPT campaign is getting increasing attention – more coverage here.
• See Stop Gen AI for a list of AI-free software – and also for support groups for those seeking to push back against gen AI
• Some good anti-AI memes – and this one seems pertinent
• The UK government backtracked on plans to make it easy for AI companies to use copyrighted works for AI training – this a victory for the artists’ campaign.

The continuing effects of the Grok scandal: UK government announced it would tighten the Online Safety Act to make sure that it includes AI chatbots. It also announced it would require platforms to remove abusive content within 48 hours once identified. The Spanish government launched an investigation into  TikTok, X, and Meta for potentially spreading AI-generated child sexual abuse material. The Indian government adopted a new rule to impose a 3-hour time limit for unlawful AI content to be removed once it is identified. Data protection watchdogs worldwide are calling for better regulation. And another law suit against Grok, including by children. And Baltimore sues xAI.

Ongoing battles over copyright: ByteDance agreed to limit use of AI video generator Seedance 2.0 after Walt Disney threatened legal action. David Greene, a public radio presenter, has a law suit against google over NotebookLM which he claims has used his voice to create an AI-generated podcast.  Recent research proving that LLMs are able to reproduce exact replicas of the texts they’ve been trained on increases the risk further of copyright law suits against AI firms. Another collective action law suit has also been levelled against Nvidia over its scraping of youtube. Nielsen’s Gracenote is suing OpenAI for copyright infringement. The London Book Fair saw people walking around with a book – Don’t Steal This Book – on display as part of a campaign by authors against gen AI and for better copyright legislation when it comes to AI training models. Related, a new report by the Society of Authors (SoA), the Independent Society of Musicians (ISM), Equity, the Association of Illustrators (AOI), and the Association of Photographers (AOP) highlights the impact of gen AI on creative workers and was used to launch a campaign that calls on the UK government to adopt new regulations.

Recent research by Social Change Lab highlights how the risk of AI-enabled warfare – autonomous weapons like robots and drones which operate without meaningful human oversight – causes the greatest level of concern about AI and therefore the most likely to prompt social movements and protest in opposition to AI. This is especially pertinent at the present, given that the Pentagon strongarmed AI firms to fall in line before the attacks on Iran. Also see a recent post by Masoumeh Iran Mansouri (one of our convenors) on exactly this question of AI for military use. The FT also published an op-ed with a discussion of the risks. The discussion surrounding the horrific murder of 175 school children, seemingly by Palantir and the US Government – in Shajareh Tayyebeh primary school in Minab, in southern Iran – is also obviously relevant here.

Palantir’s toxic reputation is beginning to have some consequences – including the Swiss government’s repeated refusal to award contracts to the firm – and British MPs urging the government to halt its latest contract with Palantir – likewise in the US politicians are distancing themselves from Palantir due to its toxic reputation.

Employees don’t like or want AI – New research based on two surveys of over 1000 employees finds that those who are using AI more are also both more anxious about the effects of AI, and more inclined to resist its introduction. More research here on reluctance by employees to take up the use of AI. Another survey saw 89% of managers report no increase in productivity. Likewise, the Economist reports productivity gains at roughly zero. Workers are also reporting detrimental mental health consequences. Even managers are now reporting that AI is a major risk. In all of this the obvious reason that AI introduction is struggling is due to human opposition. Regardless of the benefits, employees are increasingly being told they should embrace AI or otherwise suffer the wrath of their managers – Accenture released a memo telling staff they needed to embrace AI if they wanted to have a chance of promotion; google is also dialling up the pressure on its employees; and PWC threatens to get rid of any naysayers. The fact that such an authoritarian approach is needed is obviously telling in terms of how employees are reacting to AI.

Everyone hates AI slop – here’s some people on reddit complaining about its infection of Pinterest. Here’s Hoopla agreeing to remove AI slop books following a 404media investigation. This mirrors a more general trend whereby the public don’t seem so keen on AI.

Political ramifications continue: intra-Republican divisions were further exposed as the White House attacked Utah Republicans over their proposal for a new AI Transparency Act – prompting a massive billboard campaign against the White House in Utah in response. An intra-AI industry political division has also emerged with AI billionaires on both sides – a ‘safety’ AI PAC (Public First Action) backing pro-regulation candidates (in both the Democratic and Republican parties) stands in opposition to a right wing (anti-safety?) PAC (Leading the Future). Axios has a much more detailed discussion on the political divisions that AI is creating in the US. The Global AI Impact Summit in New Delhi saw various efforts by political leaders to espouse their commitment to various degrees of AI safety: Abhishek Singh, chief executive of India’s AI mission, called for “democratising AI access to the global south” and rolling out technology for social ends such as education, health and agriculture. Although US-based Big Tech firms largely dismissed all of these suggestions. Florida voted for an AI Bill of Rights. This didn’t get through – further exposing the divide within the Republican Party. The White House released its AI policy – but not at all clear this will quell the opposition within the MAGA coalition itself.

Everyonehatesdata centres. The political fallout and opposition to the widespread of data centres saw Trump seek to shift the burden/blame to Big Tech firms directly. As a partial response to the ongoing outcry about data centres across the US, Anthropic pledged to cover the additional energy costs created by data centres – subsequently witnessing most of the Big Tech firms also signing the pledge (although obviously the proof will be in the pudding on this).  A new Senate bill proposed by Josh Hawley (R-Mo.) and Richard Blumenthal (D-Conn.) sought to achieve a similar goal – although it looks unlikely to make it through the legislative process. Bernie Sanders also has a proposed bill which would pause data centres from being constructed. Chicago’s biggest electricity company, Commonwealth Edison Co. announced plans to make sure the data centres pay for the increased fuel costs. Authoritarian moves against data centre opposition also surfaced as a man was arrested for speaking too long against data centres at a city council meeting in Oklahoma. Farmers in Kentucky are refusing to sell their land to data centre developers. Data centres also continue to face refusals in planning applications, including in Pennsylvania, New Jersey, Edinburgh (despite its own city planners recommending to accept) – and a petition calling for a ban in Ohio. Campaign groups in the UK are calling for more transparency over energy use by data centres. And data centres are behind schedule (due in part to local opposition). People also don’t seem to much like AI in general.

Also some financial implications associated with data centres – Moody’s questioned the method of accounting that data centre providers are using, including special purpose vehicles, which appears to hide the level of financial risk the data centres are incurring – this is likely to become more of an issue as data centres are increasingly seeking credit ratings. Blue Owl Capital failed to secure financing for a $4 billion data center project in Pennsylvania. The CIOs are also getting worried. The FT published a discussion piece considering the possibility of a $9 trillion bust.  Guardian reports on the possibility that it’s based on ‘phantom investments’. As a result data centres are struggling to find insurance cover.

Bloomberg has a nice piece discussing the fallout from the blog post published by tech entrepreneur and investor Matt Shumer, Something Big Is Happening, which sought to highlight how AI could decimate professional jobs. As the piece in Bloomberg argues, the severe market reaction to a single blog post by Shumer reveals the fact that “AI is trading on vibes and anecdotes”. This seems to equally apply to the market fallout that followed the Citrini post.

And political contestation arising from data centre opposition – Illinois Governor JB Pritzker (Democrat) announced a two-year suspension of state tax incentives for new data center developments – with Virginia Senate proposing to do the same. Pennsylvania Gov. Josh Shapiro (Democrat) announced a plan for better regulations for data centers. More generally the Democratic Party is increasingly positioning itself as the anti-data centre party – and in the process witnessing internal divisions within the party (see the Democratic primary in North Carolina). US Senators opened an investigation into the use of energy by data centres.

Data centres were also exposed as a military target as the Iranian regime attacked Amazon data centres (see also here – and the regime’s declaration that it would continue to do so here.)

The lack of usefulness (uselessness?) of AI? – KPMG partner fined for using AI to answer an AI-training exam. A report commissioned by a group of non-profits, including Beyond Fossil Fuels and Climate Action Against Disinformation, discredited the claims of gen AI firms to have the potential to tackle the climate crisis – the report highlights how the gen AI firms promised “AI” will solve the climate crisis, failing to mention that all of such claims relate to machine learning and not generative AI/LLMs.  New research published by Anthropic(!) in which an experiment with AI-assisted coding found AI-assisted coders were marginally more productive (not statistically significant) and a significant decline (significant) in their understanding of the code itself. The UK Government’s website – Gov.uk – is using chatbots to answer citizens’ queries, but producing unpredictable errors in the process, a study found. Wired404 includes a report on Alpha School, an AI-based school which has been heralded by right wing politicians and press in the US, turns out to be generating faulty lessons which are doing more than harm than good to students. Amazon’s agentic AI, Kiros, sabotaged Amazon’s AWS – deleting and recreating the code it was supposed to be working on. Related, Amazon’s own workforce prefer not to use Kiro, but are being forced to, and resisting those instructions. A recent study finds 13% of biomedical research is AI-assisted text – plus arxiv has moved to restrict submissions of AI-generated text. European journalist – Peter Vandermeersch – was suspended over AI-generated quotes. Essex police had to pause facial recognition on the grounds of racial bias. OpenAI closed its Sora video-making app and cancelled its $1bn Disney deal – all of which seems to highlight the difficulties AI will face in replacing artists’ creative work. Now AI is also resulting in flawed polling.

Lilli, McKinsey’s AI platform used by its 40,000 staff, was hacked.

The murky world of AI billionaires was spotlighted as ‘Godfather of AGI’ Ben Goertzel was seen in recent emails to have courted Epstein for funding and congratulated him on jail release. Billionaires also had a fallout over the  Fermi America data center.

Given that ICE is powered by AI, then anti-ICE activities are a form of anti-AI activism? Wired has an excellent piece covering the different ways – analog and digital – that hackers and makers are working to provide activists with the necessary tools for anti-ICE evasion and solidarity activities.

We continue to see leaks from inside the AI industry by employees exposing the practices of the industry: OpenAI employees sought to raise the alarm regarding Jesse Van Rootselaar, who later went on to become the suspect in the mass shooting  in Tumbler Ridge, British Columbia, Canada – whilst OpenAI employees raised the alarm about Van Rootselaar’s writings, which they viewed as indicating potential for real-world violence, and sought to report this to the local authorities, nevertheless the OpenAI bosses decided not to report it- this has since been followed by a law suit against OpenAI. More generally, whistleblowers continue to be a thorn in the side of the AI industry. Workers also exposed the falsehoods used to justify the sacking of half of Block’s staff under the guise of AI-related efficiency savings – and the press ran with it.

Rest of World reports on gig workers (data labellers) in Africa working on data to train AI systems that could be being used for surveillance, including the kidnapping of Maduro – with the Data Labellers Association as one of the groups raising concerns over this. More coverage of the Data Labellers Association here.

More law suits have been brought against AI: one here claiming Gemini caused a suicide; another here against Grammarly; another here against Tesla Cybertruck over FSD crash. Meta and youtube found negligent in landmark social media case.

This tool sabotages gen AI by making it very slow

Wikipedia has banned AI generated content

Other interesting sources:

• New database on Worker Mobilizations around AI in Arts, Culture, and Media
• Rest of World covers a range of campaigners against AI

Some campaigning news:

• a new campaign group, Pull the Plug, is launched in the UK, calling for citizens’ assemblies to input into the governance of AI, and with a demonstration planned in London on 28 February – details here.
• The launch of a new boycott campaign – QuitGPT – calls for a boycott of ChatGPT, citing the fact that ChatGPT is Trump’s biggest donor – with a good story on the launch here.

Opposition from employees continues: 15,000 nurses went on strike in New York in January/February 2026, with AI as one of the strike issues – as of now it looks like an agreement has been reached in every hospital hit by the action, winning AI safeguards in their contracts for the first time – victory to the New York nurses! TMK produced an excellent podcast on healthcare workers who are members of the National Union of Healthcare Workers about their ongoing contract dispute with Kaiser Permamente. A whistleblower (former Google employee) alleged that Google helped an Israeli military contractor with AI. Employee whistleblowing also helped expose Musk’s deliberate use of porn to generate interest in Grok. There is a good summary here from Power at Work of worker responses.

Artists continue to mobilise in different ways in response to the impact of AI. Hollywood continues to be focused on limiting the impact of AI on film-making. A recent film about AI, Mercy, saw Amazon ban the use of AI – actors continue to be unhappy about the use of AI in film, and actors’ agents are including non-use of performance to make AI replicas in their contracts – and there is already talk about another strike this year once the widely-cited 2023 contracts, which famously saw a prolonged strike result in a deal to limit AI, expires. In the UK, 99% of members of actors’ union, Equity, voted to refuse to be digitally scanned on set over concerns that the scans will be used by AI. Audiences also don’t seem very keen on AI-generated content. Platforms are moving to ban AI-generated content. Bandcamp announced a ban on AI-generated songs. Comic-Con banned AI art.

On copyright, the UK goverment continues its consultation on how to legislate in the area – recent consultation saw an overwhelming rejection – 97% of submissions – of a plan to introduce a tech-friendly approach to copyright of artists’ work. India released a plan to make AI companies pay for training data. The European Parliament has called for new copyright laws.

Data centres continue to be a major focus of opposition, especially from nearby residents – in fact ‘everyone hates them’. Resident campaigns are reported in Northern Virginia, Oklahoma, Georgia (which is seeing a push for a ban), London, Buckinghamshire (UK – where the planning permission was retracted in an embarrasing U-turn by the government). Rest of World produced a good summary of environmental campaigns arising in response to the impact of data centres.

Politicians also look increasingly concerned about the unpopular rise of data centres in the US. This has prompted investigations over the impact on electricity costs. There is also growing pressure, especially – but not only – within the Democratic Party to adopt a strongly anti-AI position. Republican Governor of Florida, Ron DeSantis, has adopted a strong anti-AI stance. Internal divisions also continue to be evident within the MAGA coalition over AI – with Republican politicians concerned about local opposition amongst voters unhappy about Trump’s attempt to ban state-level regulation of AI. Indeed, despite Trump’s attempted ban on state-level regulation many states continue to move towards tighter regulation. Trump was eventually moved by this opposition to pledge to encourage Big Tech to do more to avoid pushing up electricity bills (citing Microsoft’s recent announcement on this).

The publication by Grok of sexualised deepfake images, including of children, sparked global outrage – with Apple and Google also pulled into the controversy – and a widespread call for bans (and in some cases actual bans), including in the EU (where an investigation was launched), India, the UK (also see here and here – and including the launch of an investigation by UK media regulator, Ofcom, and another one by the Information Commissioner’s Office, and proposals for legislation to ban the creation of online sexual images), Malaysia, Indonesia (where it was banned and then later ‘conditionally’ had the ban lifted), Brazil, California, and France (including a police raid) – eventually prompting Grok to turn off the image creation function, except for subscribers, which also caused a further political storm as critics claimed this amounted to allowing deepfakes of children if the Grok user paid for the service. More coverage of this here. ControlAI speaker Andrea Miotti was interviewed on Sky News to talk about the risks of AI as a result, and by a committee of the Canadian House of Commons. AOC and Paris Hilton announced their attempt to seek legislation in response in the US. The controversy is also seen as one of the reasons Spain (and later Brazil) is moving to ban social media for children (which Musk was very angry about) (prompting some social media firms agree to independent assessments of health risks), and a more general move that Rest of World claims amounts to an attempt to move off US Big Tech.

The rise in AI relationships (or human-like interactions) has also prompted concerns – China recently adopted regulation. Related, the impact of AI delusions is becoming a common theme in public discussion.

Various legal proceedings continue – Google and character.AI settled a lawsuit relating to a child’s suicide after developing a relationship with a chatbot. A lawsuit was also initiated in LA over the addictiveness of social media and its algorithms (prompting TikTok to settle).

The adoption of new regulation of AI is ongoing. Nigeria announced new plans to regulate AI. South Korea adopted strict new AI regulations. The EU’s regulatory landscape continues to be the subject of debate. The EU issued a warning to Meta over its restricted access to rival AI assistants on Whatsapp. Starlink was blocked in South Africa (and restricted in Uganda). A new angle also opened up in the debate over job displacement when a Labour minister called for UBI, paid for by tech companies, as a way of responding to the job losses associated with AI.

The AI-fuelled semiconductor chip industry is also prompting social opposition. In Taiwan, the need for largescale green energy such as through wind turbines is devastating local fishing and farming – prompting ongoing protests and opposition by those affected.

Kim Crawley, founder of the excellent StopGenAI website, is crowdfunding the publication of a forthcoming book, Technofascism Survival Guide.

The limited usefulness (un-usefulness?) of AI also continues to be exposed by a wide range of reports and surveys. One report found that half of AI projects have been shelved due to complex infrastructure. CNN reports a new trend of ‘analog lifestyles’ where people explicitly commit to removing AI from their lives. A survey in the Wall Street Journal highlights the stark difference in perception between bosses and employees – with the former heralding efficiency savings and over 90% of CEOs claiming that at least 2 hours per week is saved, yet 40% of employees claim no time has been saved at all. A report in Harvard Business Review finds AI does not reduce work; rather it makes employees work longer – which they conclude is probably unsustainable (and which has also prompted concerns about AI-related burnout. A West Midlands police chief was caught out for using AI in producing a report on football hooliganism, citing a football matches that never existed – prompting discussion of ways to avoid using AI. A majority of CEOs report zero payoff from AI spending – with very similar findings in another report. Researchers showed that AI is able to be prompted to quote copyrighted work word-for-word, despite AI firms’ denials that it would do so. DeepMind chief Demis Hassabis warned AI investment looks ‘bubble-like’. Dario Amodei, Anthropic chief, warned ‘Humanity needs to wake up’ to dangers of AI. The Register reports that ‘Only 3.3 percent of Microsoft 365 and Office 365 users who touch Copilot Chat actually pay for it, an awkward figure that landed alongside Microsoft’s $37.5 billion quarterly AI splurge and its insistence that the payoff is coming’. The announcement of a $660b investment in AI infrastructure by the Big Tech firms was not received well by the markets. There was an interesting lengthy discussion in Bloomberg on whether robotaxis are safer than human-driven ones – in sum, at least in terms of deaths per miles driven the answer is no, and otherwise it is that we don’t know. Another report discusses the rise in hiring of good writers, due to the poor quality of AI slop writing (or ‘slopaganda’).

The link between the ICE raids in the US and AI has been getting increasing attention – prompting French company CapGemini to drop a contract with ICE following pressure from French officials (and then moving to sell the unit altogether). Palantir and Roshel have also faced scrutiny.

The use of AI by organisations and firms remains controversial. Campaigners highlighted the use of AI to write the political platforms of far-right parties during the 2024 European Parliament elections. The British Museum faced criticism for using AI-created content.

In terms of detrimental impact: Campaigners highlighted the negative impact of AI when used to teach children in schools. Opposition is also growing over the poor health advice that AI is producing – with Google removing some of its summaries. MPs in the UK are highlighting the risks of unregulated AI for finance. The Washington Post covered a story on academics increasingly turning to oral examinations as an alternative form of assessment to deal with the difficulties of students using ChatGPT. Concerns are also increasingly being raised regarding the impact of AI use on younger employees’ capacity to learn.

Many critics look to open source alternatives. This article discusses the use of open source mapping platform, OpenStreetMap, used by Palestinians to navigate roads which they are prohibited from using by the Israeli state in the West Bank.

An attempt to stage a pro-billionaire march in San Fransisco flopped spectacularly.

Data centres continue to be a major target of public opposition, especially in the US, and over the pollution for local residents and water and electricity use. Amazon was under pressure after being found to have actively sought to keep the water use of its data centres secret. As Harvard Business Review reports, some state agencies have begun conducting health impact assessments of data centers’ on-site diesel generators to inform emission limits. Bloomberg carry a report claiming that electricity prices have doubled since 2020, with the highest rate of increase in locations that are within 50 miles of a data centre – all of which is creating considerable public anger. Elections are increasingly focused on candidates’ positions on data centres: a recent election in Northern Virginia witnessing both candidates pledge to block the expansion of data centres, and with the vote seemingly going to the candidate who was most ardent in his opposition. This comes amid reports of America’s coming war on data centres here. The campaign organisation, Data Centre Watch, produced a report highlighting how $64 billion in U.S. data center projects have been blocked or delayed by a growing wave of local, bipartisan opposition since 2024. Also see more details of the ongoing campaign by the  Pyramid Lake Paiute Tribe to oppose the building of data centres in the region. 230 environmental groups demanded a national moratorium on new datacenters in the US. This is alongside a new online campaign to oppose the building of data centres in Pennsylvania. At the same time, Amazon has complained to the the Public Utility Commission of Oregon alleging that Portland-based PacifiCorp is failing to provide enough electricity to its data centre! 

Worker and employee resistance and opposition to AI also continues. Ex-AI product safety lead, Steven Adler, broke ranks with the firm to highlight the continued risk of serious mental health issues associated with the use of AI chatbots. Likewise, the Guardian reports from a range of AI workers and raters who report the flawed models of AI training and speak out against the use of AI. For a broader discussion, see this Podcast by two founders of a new mutual aid and advocacy group called Stop Gen AI, which formed this year out of the critical need to provide material support for creatives, knowledge workers, and anyone else impacted by generative AI.  Research by Deloitte found that employee’s trust in their employer’s AI tools is dramatically falling – by 31% in the most recent findings. Quartz reports widespread hatred towards MS Copilot, in part due to employers forcing its adoption upon employees, prompting widespread online mockery and reports that MS is now needing to downgrade sales expectations.

The impact of AI on worker productivity has also been questioned. Research into the use of AI by software developers found that it made them 19% slower than when they worked without AI – yet despite this they believed that AI had sped them up by 20%.

AI has also faced ongoing legal challenges. In Japan, Yomiuri Shimbun and the Asahi Shimbun have filed a lawsuit with the Tokyo District Court in August against U.S. AI startup Perplexity AI; Amazon also sued Perplexity in an attempt to stop its chatbot from shopping on its platform. OpenAI was sued by firm Cameo, with OpenAI accused of deliberately confusing consumers by introducing a new “Cameo” feature on Sora; a series of lawsuits by OpenAI chatbot users for suicides and harmful delusions – confirming media reports of a trend of mental health hospitalizations as a result of chatbot use. A lawsuit was filed in California against discriminatory recruitment practices as a result of AI decision-making. A lawsuit was brought against OpenAI claims ChatGPT persuaded a user to kill his mother and himself. The European Commission announced an investigation into Google’s AI overview feature.

AI firms are being forced to act as a result of these legal challenges. For instance, the fallout from an earlier adverse legal ruling against it prompted Character.AI to move to ban children under 18 from using its chatbots.

Not only AI firms, but also AI users have faced legal challenges. Most notably, courts in Utah, Indiana and California have fined lawyers for using AI-generated legal submissions that included hallucinations reporting non-existent research and case law. 

In terms of AI regulations, family-led campaigns over the risks of AI chatbots have featured heavily in recent weeks. Senators Hawley and Blumenthal have proposed legislation passing through US Congress seeks to limit availability of chatbots to minors, following pressure from parents of children who have taken their own lives under the influence of chatbots. The UK government also announced that it is considering new laws regulating the use of AI chatbots by children. In addition, China adopted regulation designed to restrict fake images and videos, adding to existing legislation which requires AI-related companies to use generative AI services in line with the values of socialism (and which already governs the design and use of DeepSeek). A new law in New York regulates the use of AI to develop personalised pricing. The  Data (Use and Access) Act 2025 was adopted in the UK, banning deepfake porn and making it a criminal offence, prompted by a feminist campaign group make up of victims of deepfake porn and experts. 

As the UK Government continues to consider AI-related copyright law, Paul McCartney contributed to ongoing objections from the music industry regarding the infringement of copyright by AI firms, releasing a track of an almost completely silent recording studio, to highlight what the music industry might become if copyright laws aren’t updated to take account of AI.

Schools and parents are becoming increasingly focused on the risks of AI, with a parent-led campaign creating pressures on Malden Public Schools and demanding better regulation, with many parents calling for an outright ban in schools.

A broader anti-AI sentiment can also be discerned in terms of public opinion and how this is translating into political pressure, and forcing firms and organisations to reconsider their approach to AI. Research into the degree to which the public trusted AI to allow politicians to take decision was overwhelmingly opposed in both the UK and Japan. Google was forced to remove its  AI model Gemma from its Studio platform after a Republican senator said it “fabricated serious criminal allegations” against her. The academic paper publishing platform, arXiv, announced it will no longer accept computer science review articles and position papers due to the overwhelmingly high level of “AI slop” that was being submitted to the platform. McDonald’s faced online opposition to its AI-generated Christmas advert, forcing the firm to turn off comments on YouTube, before setting the video to private.

In terms of political consequences, divisions within the MAGA camp have emerged following Trump’s backing for a plan to restrict state-level regulation of AI, with Ron DeSantis calling Trump’s plan “an insult to voters” (another similar report in the Washington Post). As well as divisions within Meta, with reports of a schism between the AI team and the rest.

A number of other campaign developments: The Verge reports a campaign seeking to influence the Pope to speak out against the risks of AGI. A detailed set of recommendations for campaign organisations seeking to use social media platforms in a way that outsmarts algorithms designed to minimise the visibility of Palestine solidarity campaigns. The campaign by ControlAI to ask UK-based MPs to sign a pledge calling for “binding regulation on the most powerful AI systems” reached 100 MPs signed up.

The European Conference on AI (ECAI) which took place in Bologna in October 2025 witnessed a demonstration staged by local Palestine solidarity activists and organisations, which took place outside the Conference, declaring that such conferences are ‘just advertising stores for criminal multinational companies that finance the development of new technologies, devices placed on the market with the label “for civil use”, and then sold to the “defense” sector, to companies like Leonardo S.p.a., that still export weapons to “Israel”. The production of genocide begins in European classrooms, laboratories and research centers: ECAI is the emblem of neoliberal hypocrisy, the same one that tries to hide death and destruction behind empty principles of peace, freedom and democracy’.

For more summaries of contemporary anti-AI campaigns, see this nice article in Vox listing some of the most impressive recent achievements for anti-AI campaigns. And another similarly good overview of some recent campaign successes in this article in MROnline. 

At the micro-level we continue to see an ongoing series of everyday forms of ‘misbehaviour’ or subversion regarding the use of AI. A recent survey of company employees found that over 30% were actively sabotaging their company’s AI strategy, with the most likely reason given that the introduction of AI amounted to an explicit attempt to subvert the attempt to replace employee’s jobs. Reflecting this trend, a recent McKinsey report that while 80% of companies are introducing gen AI in some form, at the same time around 80% also report no improvement on their bottom line as a result of introducing AI; the report identifies a key reason as “implicit resistance from business teams and middle management due to fear of disruption, uncertainty around job impact, and lack of familiarity with the technology.”. Likewise, the NYT reported on attempts by job applicants to trick chatbot recruiters for job applications. The Information reports on how coders are refusing to use AI tools to code, with two software engineers at Mixus, San Francisco, staging a rebellion by refusing to follow instructions to rely on AI coding-assistance software. Higher up the hierarchy within AI companies, the FT reported an exodus from disgruntled senior staff from Musk’s xAI. Also as a result of worker opposition, Microsoft announced it would disable some services to Israel’s Defense Ministry, after a company review concluded that Israel was using Microsoft’s cloud storage services to hold surveillance data on Palestinians – this came after disgruntled employees entered company president, Brad Smith’s office, to protest, hang banners and occupy rooms in the company HQ.

In terms of regulation, we have seen a number of key developments. This includes the introduction of the new California AI Companion Law (SB243), as this piece from James Muldoon reports. Newsom also signed the first AI Safety Law (SB53) in California – as also reported in more detail here by techcrunch- although he also vetoed another piece of legislation (AB1064) that sought to restrict access to AI for children. An AI Risk Evaluation Act was introduced into the US Senate. The much-vaunted UK’s AI Bill also still remains waiting in the pipeline. Calls for further regulations were also voiced. Perhaps most visibly, the Global Call for AI Red Lines saw over 300 prominent figures call for governments to reach international agreements regarding red lines for AI. Similarly, a group of leading experts in Cognitive Sciences signed an Open Letter called for the rejection of the uncritical introduction of AI in academia. For a more global summary of AI policies and regulations, do see the excellent new Artificial Intelligence Policy Observatory for the World of Work (AIPOWW) Symposium in the current issue of the journal, Global Political Economy. This reports on the work of the Artificial Intelligence Policy Observatory for the World of Work, based at University of Essex’s Centre for Commons Organising, Values Equalities and Resilience and which is led by Phoebe V. Moore and Peter Bloom (both members of our Contentious Politics Of AI Network). It includes excellent commentaries on AI policies and regulations in the EU, Brazil, India, China, and Canada.

The wider impact of AI-related infrastructure is also being contested. Efforts to oppose the building of data centres are growing in prominence. Earlier this year saw reports of the Sierra Club seeking to overturn the licence for a data centre as part of their campaign in Reno, Nevada. MIT Technology Review covered a related campaign by the Pyramid Lake Paiute Tribe to tackle the use of water by data centres that threatens to divert it away from the Pyramid Lake. More generally, the impact of AI on energy costs is starting to have political consequences. With voters increasingly calling for political measures to offset or reduce the impact of AI energy use on household energy costs. In the US, Democrats, led by Elizabeth Warren, have sought to politicise the issue and called for Trump to act. A recent election in Virginia saw both the Democrat and Republican candidates cite the rise in electricity costs as a key reason why the building of any new data centres should be blocked. This reflects a more general trend whereby public opinion in the US is becoming increasingly sceptical about the merits of AI. As the Atlantic reports, the ongoing negative public opinion of Musk continues to have a detrimental impact on the introduction of AI technologies with Tesla.

Finally, in terms of legal challenges, Anthropic agreed a much-reported $1.5billion copyright settlement to resolve the copyright class action brought against it.