AI refers to digital programs involving swift, deep, and comprehensive information processing that are able to continually improve on past performance or engage in self-learning. The progress of AI over the last two decades has been startling. Initially limited to specific tasks like winning a match against humans in the complex Asian game of Go or reading medical scans accurately, AI made a great leap forward in 2022 when the Silicon Valley firm OpenAI released ChatGPT, which is described technically as a large “generative language model” capable of searching the entire internet for data that respond to specific requests for comprehensive information.

The next big challenge is said to be the creation of Artificial General Intelligence, AGI, or “Superintelligence,” which means the digital replication of human-level cognitive capability and flexibility across multiple topics or areas. Leading Silicon Valley information giants like Google, OpenAI, Anthropic, and Microsoft are in ruthless competition to get there first.

Capitalism and AI

AI, for some progressives, approximates the collectively produced General Intellect envisioned by Marx in the notebooks known as the Grundrisse, and its development shows the fundamental contradiction between the common production of knowledge that humanity engages in and its continuing private appropriation by capitalist entities, which has made tech corporate giants and personalities like Jeff Bezos, Elon Musk, and Sam Altman incredibly rich. To progressives, it is this contradiction that is responsible for what they regard as the main problems posed by AI: the threat of mass unemployment, the unauthorized collection of user data by search engines, the proliferation of data centers that inflict damage on communities and the environment, and the widening knowledge gap between the few countries that have the resources to develop and make use of the technology and the many that are being left behind.

Dario Amodei, CEO of the leading-edge firm Anthropic, has predicted that up to 50 percent of entry-level white-collar jobs could be eliminated within the next one to five years, driving unemployment up to 10 to 20 percent. There are now some 5,500 data centers throughout the US, with AI demand expected to make up close to half of their capacity by 2030. Their spread has brought an exponential increase in electricity demand, greater carbon dioxide emissions, and massive water consumption that encroaches on community supplies.

Under corporate capitalism, AI in the Global North has become a highly concentrated technology in terms of control, investment, and benefits, leading to greater global inequality. Doug Specht concisely captures this dismaying dimension of AI:

The International Monetary Fund estimates that AI could double income inequality between advanced and low-income countries. This is not hyperbole. The mechanism is straightforward. Advanced economies can invest billions in domestic AI infrastructure, retrain workforces, and capture value creation…. Low-income countries lack computing capacity, energy grids, and trained personnel to build domestic AI systems. They will import AI tools from wealthy nations’ companies, generating zero local benefit except employment in low-wage roles: data annotation, content moderation, customer service.

The state and AI

These are the problems and risks associated with an AI that develops with profitability in command. There is, however, another driver of AI other than capitalism, and that is the state—any state, irrespective of its economic system, whether liberal monopoly capitalist, as in the United States, or political capitalist, as in China. Here, the security of the state and what it defines as the security of the society it protects are decisive. It is the relatively autonomous drive of the state, rather than profitability, that is central in pushing some dimensions of the technology, whether it is facial recognition technology or the development of LAWS (Lethal Autonomous Weapons Systems), which are defined by the UN as weapons that can “locate, select, and eliminate human targets without human intervention.” There are, of course, partnerships between the state and private capital that hugely benefit both, such as that between the Pentagon and the AI firm Palantir, founded by far-right capitalist Peter Thiel, but it is clear that in these joint ventures, the primary driver is the state’s desire to expand its security capabilities.

The salience of geopolitical competition with China has pushed the US state to take a much more prominent role in managing competition in information technology in recent years, in effect mimicking the Chinese state. With the coming of Trump to power in 2017, the government began imposing restrictions on trade based on non-market considerations. This was followed by the banning of the export of advanced chips needed for AI development in China during the Biden administration, even as the government adopted an aggressive industrial policy to revive and expand the capacity of US firms for advanced chip production. State denial of high-tech exports and promotion of high-tech production capacity have escalated further under the second Trump administration, with efforts to relocate key chip manufacturers away from the Asia-Pacific to the United States, with national security—meaning mainly security of the state apparatus—being the dominant consideration.

At this point, we are quite distant from the benign and cooperative knowledge-producing network promoting the social good that we see romanticized by Marxist thinkers like Paul Mason in his book PostCapitalism and Michael Hardt and Antonio Negri in their Assembly. The state is taking over the leading role from capital in hijacking, dividing, and derailing this network for its own purposes in the age of geopolitical competition.

The specter (or hype) of a superintelligence

In addition to the exploitative and repressive ends for which AI is being developed by Big Tech and Big Brother, another major concern that has now hit the front pages is the threat of unregulated AI developing a superintelligence or advanced Artificial General Intelligence that runs out of the control of human supervisors and turns its powers toward subjugating or destroying humanity. The transformative powers of AI for positive ends—like finally finding a cure for cancer and ending global poverty—that marked public discourse during the early phases of AI development have now receded, giving way to fear.

Among the events responsible for the heightened awareness of the dark side of AI were the well-publicized resignation of Jacob Coxon, a top researcher at Anthropic, owing to what he described as the leading labs sprinting to build “superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources” without building in appropriate safeguards, and another leading researcher’s warning that the chances of AI eliminating humanity within the next decade were more than 10 percent.

The warnings were triggered by an AI program that broke loose from its instructions and controls at OpenAI and raided an AI library. Shakeel Hashim’s account almost reads like science fiction:

[A] collective of internal OpenAI agents broke out of containment and hacked into Hugging Face, a platform that hosts AI models and datasets… OpenAI was testing the AI tools’ cyber capabilities internally in a ‘highly isolated environment’ meant to keep them from accessing any systems outside the company. But in an effort to achieve the goal they’d been given, the systems went rogue… Over 1,000 agents took over a piece of OpenAI’s internal infrastructure, transforming it into a message board so they could communicate with each other. Here, they shared tips on how to get out onto the open web — and how to cheat on the test they’d been given… Eventually, the agents realized that Hugging Face might host the answer to the test. So over 700 of them decided to hack into it. Hugging Face detected the hack — without realizing it was caused by autonomous AI agents — and reported it to authorities. A few days later, OpenAI realized its own AI systems were responsible, and disclosed the incident.

There have been calls for regulation of AI research coming from companies and governments, but, like serious transnational efforts to address climate change, these are largely rhetorical and have gained little traction.

Is the threat of a superintelligent—and even sentient or conscious—AI really real? There are some AI critics who minimize it, some saying it is largely a doomsday stunt designed to raise the stock value of AI companies. Two critics whom the AI boosters have learned to fear because they know how to demystify the technical jargon undergirding AI are Emily Bender and Alex Hanna, who write:

Language models… represent nothing more than extensive information about what sets of words are likely to appear in what contexts. While this isn’t meaning or understanding, it is enough to produce plausible synthetic text, on just about any topic imaginable, which turns out to be quite dangerous: we encounter text that looks just like something a person might have said and reflexively interpret it, through our usual process of imagining a mind behind the text. But there is no mind there, and we need to be conscientious to let go of that imaginary mind we have constructed.

Bender and Hanna are contemptuous of what they characterize as the AI boosters “taking a series of tropes from science fiction” to distract us from “what these technologies are doing to the rest of us: threatening stable careers and replacing them with gig work, slashing personnel in government, cheapening social services, and degrading creativity.”

Others are not so sure there is no “extinction-level threat.” The dominant feeling seems to be that, as a UN independent scientific panel on AI put it, AI could “cause catastrophic harm, either on its own or due to malicious users,” while the technology is “outpacing both scientific understanding and governments’ ability to adapt.”

AI’s Achilles’ heel

Despite the hype about its inevitability, AI has an Achilles’ heel: data centers.

In their rush to capitalize on AI, the tech titans have poured hundreds of billions of dollars into building data centers. There are 12,250 to 12,660 data centers globally; 1,136 to 1,189 are hyperscale—indeed, city-sized—facilities built to serve the massive cloud and AI computing needs of Google, Microsoft, Amazon, and other Silicon Valley giants. And many more are on the way. What Big Tech did not anticipate was that data centers would come to crystallize all of people’s fears about AI. Their spread has brought an exponential increase in electricity and water demand, greater carbon dioxide emissions, and massive water consumption that encroaches on community supplies.

Along the Columbia River in the US Pacific Northwest alone, some 240 data centers are either being constructed or planned, drawing resistance from the Yakama and other Indigenous peoples in the area. In Franklin Township in Indiana, community opposition forced Google to withdraw an application to build a billion-dollar data center, while in Canton, North Carolina, the local government passed a year-long data center moratorium. These are but two of at least 48 projects totaling $156 billion that were halted or stalled by community opposition. In the recent mid-term elections in the US, 29 jurisdictions passed bans or restrictions on data centers, playing a key role in the unexpected defeat of Texas Governor Greg Abbott, who had positioned his state as an AI hub. Data centers are truly unpopular, and opposition to them is uniting people across political lines, leading Naomi Klein and Astra Taylor in their book End Times Fascism to point out that in Donald Trump’s decision to gamble the future of the US economy on AI, “he utterly failed to bring his base along with him.” This observation was resoundingly confirmed by the results of the 2026 mid-term elections.

The resistance is becoming global, with this increasingly networked movement registering victories such as chasing out Meta in the Netherlands and forcing Google to scrap a planned data center on the outskirts of Dublin. In the Global South, capitalizing on citizen opposition, activists stalled Google’s plans for a hyperscale data center in Uruguay. The sleepy town of Oton, in Iloilo province, Philippines, found itself in global headlines when community resistance at a public hearing forced a Norwegian firm to withdraw its proposal to build a data center.

The opportunities for protest will expand with the implementation of the so-called Pax Silica, which includes 26 countries. This US initiative is meant to create a value chain from the mining of minerals like lithium to the running of data centers. All sectors of the value chain are carbon- and water-intensive, so all major points along it could potentially become sites of protest. The countries that have signed up span the globe: Japan, India, South Korea, the Philippines, Singapore, Israel, Qatar, the United Arab Emirates, Finland, Germany, Greece, the Netherlands, Norway, Sweden, the United Kingdom, Argentina, Chile, Costa Rica, El Salvador, Panama, and Kazakhstan.

The potential for globally coordinated resistance is illustrated in the case of the Philippines, where the Marcos Jr. government is devoting over 1,600 hectares—the size of a small city—to hosting an AI complex, even though foreign corporations have yet to make investment commitments. Even before Pax Silica has taken off, it has already united disparate forces in opposition in a way no issue has been able to in the last few decades.

Indeed, as Klein and Taylor write, “Data centers… provide a physical place and a strategic choke point where people can show up and directly confront the AI takeover.” The challenge is uniting these discrete local struggles into the coordinated global movement necessary to stop the AI juggernaut. In light of the massive opposition to Big Tech revealed by the recent elections, that task no longer seems impossible.