In the crosshairs of AI (part two)

In their mad rush for profit and power, the tech oligarchs have unleashed artificial intelligence (AI) into nearly every corner of life on this planet. Firms are investing billions in data centers, which house the infrastructure to process the enormous amounts of information that power AI. They are also plowing ever more money into research and development. Their hope is that the technology will overcome global capitalism’s slump, which the system has been trapped in since the Great Recession, and trigger a new epoch of expansion. States are incorporating it into their police forces for domestic surveillance as well as their militaries for cyberwarfare and high-tech weapons. Ominously, the struggle for AI supremacy is now at the center of the imperialist rivalry between the U.S. and China.

The widespread, unregulated use of AI is already having devastating impacts on our environment, jobs, and rights. Some worry that it threatens our very species with extinction. In this three-part interview, Tempest Collective’s Ashley Smith asks two advanced tech workers, Josh and Yasmine, about the technology, its much hyped promise and threat, and its impact on our lives. Most importantly, they lay out how the Left and labor movement should respond and what we should demand and fight for.

Part Two: Plunder in the Global South, Labor Resistance, and Imperial Rivalries

Ashley Smith: Karen Hao, in her path-breaking book Empire of AI, has exposed the real material impact of AI in the here and now–its brutal exploitation of workers in the Global South, its fueling of extractivism there, and the massive impact everywhere of data center construction on the environment and the cost of energy. Can you talk about AI’s immediate threat in these various dimensions?

Josh: It is an incredible book; everyone should read it. Karen Hao documents the labor involved in labeling data to train AI, and in judging and editing responses to user prompts to make them more appropriate for chat situations. This is a process called Reinforcement Learning by Human Feedback.

This labor is often extremely traumatic. The workers view graphic sexual and violent content and classify images and text to train the models to categorize and moderate content. It is terribly paid and particularly precarious labor. Tech companies subcontract to firms that employ these workers, mainly in the Global South. When these outfits move their operations from one country to another, they leave a trail of broken families, psychological trauma, and impoverished people. But Hao also documents the workers’ resistance and struggle and some successes in regulating and improving their pay and working conditions.

Mainstream accounts of AI rarely see or discuss this labor, and Hao corrects this by making it a centerpiece of her story about OpenAI’s growth and success as the main pioneer of this new technology.

As you point out, she also connects these workers’ struggles to resistance to building hyperscale data centers worldwide. Whether in rich or poor countries, corporations are mainly setting these up in poor and semi-rural communities.

In their race to build these data centers, corporations have secured environmental permits with dubious oversight and barely any regulation. Instead of using renewable energy, they rely on generators powered mostly by dangerous, polluting fossil fuels. To cool the massive data processors, the material basis of computing power, these centers consume massive amounts of water, often at the expense of agriculture and people.

Resistance to data center construction has become a central point of struggle against the expansion of AI, and a point of leverage for working people to demand democratic control over our lives and environment. Hao documents some of the networks emerging from this resistance and potential approaches to creating more equitable and sustainable data centers, as well as effective ways to stop them from being built at all.

The introduction of AI has had, and will continue to have, far-reaching and often devastating impacts on workers and the environment. AI is also becoming a key technology for domestic and international surveillance and war. This is an opportune moment to tie these struggles for democratic control of our lives together across industries, regions, countries, and workplaces.

Yasmine: It would be hard to overstate the climate impact of the AI boom. In both the US and China, fossil fuels are the primary source of power for data center expansion—specifically natural gas and coal, respectively. 

Intense capitalist competition over AI development has created urgent energy and water demands that are incompatible with a transition to renewables and efforts to mitigate the effects of global heating, like drought and dwindling drinking water supplies. The IEA estimates data centers “will account for at least 20 percent of the rich world’s growth in electricity demand by the end of the decade.”

In the US, big tech companies have all thrown their climate pledges out the window. And they’re building data centers that guzzle trillions of gallons of fresh water (in places like Texas, Arizona, and California!) in the midst of record hot temperatures, drought, and wildfires. All the while, they maintain the sales pitch that AI will enable us to solve climate change.

The race for AI supremacy is also driving intensified extractivism with mounting environmental and social costs. Companies have increased mining of rare earth minerals in the world’s most exploited and impoverished regions to make magnets for Graphics Processing Units (GPUs), which are the infrastructure for LLMs.

Production of these minerals can be extremely water-intensive (accounting for up to 65 percent of water use in some regions) and produces toxic waste that poisons water sources and farmland. Workers in these regions, from Chile to the Democratic Republic of Congo, are hyper-exploited, and these communities suffer severe health consequences due to long-term exposure to heavy metals.

The use of these technologies in military and policing applications has turbocharged the expansion of the surveillance state and produced the world’s first AI-assisted genocide in Palestine. High-tech companies have enabled the collection and analysis of thousands upon thousands of terabytes of data, the real-time tracking and surveillance of civilians, and the generation of thousands of “targets” for bombardment.

The use of AI has amplified racism in the criminal (in)justice system, something that organizations like the Algorithmic Justice League have been raising the alarm about for many years. ICE agents are actively using facial recognition software to track and kidnap our neighbors. Activists and civilians engaged in any kind of protest or civil disobedience have taken to covering our faces in marches and at actions, recognizing the state’s expanded ability to spot, track, and target each of us.

The pollution of our information ecosystem is another consequence that I think we’ll see play out in the coming years. For example, we’re seeing scientific journals defer to AI reviews over their actual editorial boards and, as a result, publish erroneous research. Deepfakes—AI-generated images and videos of real people doing things they didn’t actually do—have enabled a new scale of sexual exploitation and abuse of women and children. An estimated 98 percent of deepfakes are pornographic, and 99 percent of those involve women.

Unfortunately, I could go on listing the damage already caused by the AI boom. This technology has been developed and inserted into our lives without our consent or any democratic input from workers or communities, to the detriment of our safety, health, and the prospects of this planet’s survival.

AS: AI’s development has also led to an enormous concern about its impact on workers’ jobs. Some have predicted the elimination of whole swathes of jobs. How is AI impacting jobs now? Is the panic about elimination of jobs warranted? How is this similar and different from other technological breakthroughs in capitalism?

Josh: Cory Doctorow has written one of the best books about this. It’s called The Reverse Centaur’s Guide to Life after AI. The centaur has been used as a metaphor for the relationship between a person and AI: the human drives the relationship and uses the technology to improve and augment their abilities. Doctorow contends that the opposite is actually the case. AI algorithms are, in fact, guiding human labor to exploit it more effectively. He offers a gut-wrenching, enraging account of Uber algorithms using gamification techniques to get drivers to accept rides at low commission rates, completely decoupled from the surge prices a rider might be paying. AI is being deployed as an avatar for capital. This is a reverse centaur dynamic. We are no longer at the helm.

In Mute Compulsion, Soren Mao shows how capitalist class relations exacerbate this dynamic. The system’s competitive drive compels bosses to compete with one another, increasing worker exploitation to generate more growth and profit. AI has been created and trained in the world of capitalist relations, in which the law of accumulation for the sake of accumulation governs AI.

This doesn’t mean that AI will or can take over our jobs. Just because our bosses hope for that outcome doesn’t mean it will be. And Doctorow, among others, is skeptical they will succeed, though he quickly points out that if they want a reason to fire you, bosses will use AI as an excuse.

There are many factors to consider here, some of which I have touched on already. One of them is the cost of human labor versus machine labor. We have been able to automate grape picking for years, but it is still cheaper to get humans to do it. That said, if the bosses do replace workers with machines like AI, they will likely drive down the remaining workers’ wages.

But the ability to replace or reduce human labor is not as simple as frontier companies advertise and imagine. Even simple tasks involve tacit, nuanced knowledge and behaviors that a model can’t capture or train AI to do. Humans are the connective tissue that binds PDFs, spreadsheets, phone calls, and judgment calls about other conflicting tasks to data recorded in various systems.

Putting AI into this messy mix requires thoughtful labor and a transformation of existing workflows. This is being done clumsily, with high expectations and low effort, leaving workers with yet another point of contact to input, consume, and correct the output, with no clear benefit. Just as AI is a jagged technology, our workplaces and processes are also uneven and unevenly integrated. Using AI to increase productivity will take time and effort.

I think we can see AI as a subset of, and an extension of, the electrification of our work and personal lives. It took capitalists a long time to realize the economic benefits of electrification, but today no one can deny that it was a significant shift. I think AI will likely be even more significant than electrification, digitalization, and “internetification” of labor and life.

The pace is hard to predict; the challenges I laid out above are real, but development is accelerating, so I am not sure how to predict this without more thought and research. One sobering reference might be the fax machine, which, despite being technologically redundant as email became widespread, persisted as a common tool in US healthcare into the 2020s.

Regardless of the timing and the likely intervening booms and busts, capital can and will use this technology, even with its current limitations, to significantly increase workers’ productivity. Those who agree with this predict different impacts on labor. Some say the new workflows will require new job roles and enable more production, which might unlock more latent demand in a dynamic known as the Jevons paradox: more productive technology leads to more jobs. On the other hand, others argue AI will lead to mass layoffs. Both may be true across time, geographies, and industries.

We cannot rely on aphorisms and historical anecdotes to predict how this will play out. Working-class organizing and the struggle for democratic control of technology will, as always, be a decisive factor in determining its impact. In the end, the class struggle will determine the results, whether in favor of the tech oligarchs or us.

AS: Tech workers have been organizing over the last several years about the use of their tech in support of US imperialism and Israel’s genocidal war in Palestine as well as over their own laboring conditions. How are tech workers in particular threatened by AI? How are workers responding to it?

Josh: Perhaps the largest and most significant uses of AI are for war and surveillance. It’s central to everything from Flock cameras to autonomous drones, autonomous tanks, spying and search algorithms, and demographic data analysis. These applications lead to lucrative contracts for the tech companies and, despite their utopian mission statements, capitalist competition makes it a business imperative to ink deals rather than leave money on the table.

This is often at odds with the ethical and political positions of the tech workers at these companies who have organized to prevent these contracts. We have staged some significant actions like the Google walkouts and scored some victories. But there have been more losses.

The mostly un-unionized workforce has shown incredible solidarity and taken bold action to assert democratic control over the technology we work on. The tech companies took a while to figure out how to react, but now they’ve settled on age-old tactics: firing the most outspoken first, then following up with more layoffs to send a chill through the workforce.

I think it has had some effect. I am scared to organize, and to be honest, I am not sure what the best strategies and tactics are right now. I think we need to decide this collectively and through practice (and not necessarily the same ones we have tried up to this point).

Timing will be everything, and acts of solidarity will be key. Perhaps industry-wide, rather than company-specific, organizing and demands for industry regulation will be most effective? Or maybe just incrementally building networks amongst the rank and file?

Yasmine: Over the past several years, tech workers’ leverage in the labor market has eroded pretty dramatically. Every year since 2020 has seen mass layoffs in the tech sector, except in 2021. Layoffs peaked in 2023, with 265k employees laid off, followed by 120k to 150k laid off each year since. Tech companies only started attributing mass layoffs to AI in the last couple of years.

Mass layoffs, coupled with hiring freezes, have meant many more workers are competing for each open job, a dramatic shift from years past, when tech companies competed with each other for talent. Company culture, compensation, and benefits were key tools tech companies used to attract talent, and it was standard practice for tech workers to jump between companies for higher pay.

I think it’s safe to say those dynamics have largely disappeared for most tech workers. Tech salaries decreased by 9 percent between 2022 and 2023, while tech executives’ pay has gone up 45 percent between 2021 and 2025.

Although tech companies are betting on future productivity gains from AI tooling, I think the layoffs have primarily been driven by AI investment, not the presumed productivity gains. Mass layoffs are a cost-reduction measure happening in tandem with large AI infrastructure investments that haven’t yet turned a profit. A Meta internal memo makes this explicit, stating that the company’s cuts in headcount were made specifically to “offset the other investments we’re making.”

Given all of this, I think it’s important to press that, at least thus far, the productivity gains in tech are still projected/theoretical. AI has been used as an ideological tool to send the message to tech workers that we’re becoming redundant, making us more amenable to decreasing pay and worse working conditions where we’re taking on the labor that used to be done by our laid-off colleagues and in which we’re expected to use AI to increase our output tenfold. The previously controversial 9-9-6 model, a 9 a.m. to 9 p.m., 6-day-a-week work schedule, is today becoming more commonplace in the tech job market. 

But there is a silver lining. These declining conditions and the increasing precarity are laying to rest the mythology of tech as a uniquely fair meritocracy and tech workers as somehow exempt from the rules of capitalist exploitation. More and more tech workers are beginning to identify with other workers rather than with the tech boss, which many used to aspire to become.

This is happening along with the cultural shifts forced by the organizing and sacrifices made by tech workers taking action to push these companies to end their participation in Israel’s genocide. These actions show that confronting bosses is necessary to win demands and that companies’ professed commitment to ethical conduct has never been grounded in structure and was never enforceable without worker action. These are some of the most inspiring examples of resistance:

• The No Azure for Apartheid campaign forced Microsoft to partially cut ties with Israel by organizing workplace actions, petition campaigns, protests at executives’ homes, and more. After years of struggle, the campaign achieved one of our few victories, forcing the company to cut a contract with Israel’s criminal unit 8200.

• Kick Starter Workers Union organized a 45-day strike last year, during which they won AI protections among other demands.
• Google DeepMind workers in the UK and Ireland voted to unionize earlier this year, campaigning primarily on the issue of AI use in military applications and human rights abuses.

I think we’re just seeing the beginning of possibilities of a tech-worker rebellion. The task is especially daunting at large hyperscaler companies, but our conditions will make confrontation inevitable, and I don’t think we have any other option. I’ll also take this opportunity to plug the new Haymarket book, Against Tech Oligarchy, by Clarissa Redwine and JS Tan, which does an excellent job examining the history of tech worker organizing over the last decade.

AS: The race for AI dominance is not only between capitalist firms within the US but also between states, particularly the US and China. Indeed, it has begun to resemble the nuclear arms race between the US and the Soviet Union. The Trump administration has used this as an alibi for blocking any regulation of the industry, claiming that whoever wins the race ensures dominance over global capitalism. How advanced is China compared to the US? Why is the US so worried about Chinese competition? How does the race between the two states compare to the nuclear arms race during the Cold War?

Josh: Dario Amodei, the CEO of Anthropic, is a hawk in this regard. In his recent essay, he warned that any regulated slowdown or pacing of AI development in the US risked cutting into “the lead that US companies have over authoritarian regimes, chiefly the Chinese Communist Party. If we slow down by more than this amount, then (unpaced) CCP-associated projects will pull ahead, creating significant national security risk.” With no sense of irony, he is whipping up a panic about authoritarian China at the very same time that Trump is undermining what passes as democracy in the US and turning it into an authoritarian, undemocratic, and violent state in its own right.

Nevertheless, the imperialist competition between China and the US, as well as lesser powers, is very real, and the role of AI in these rivalries is not just hype. This is one of the real dangers of people on the Left downplaying the technology’s capabilities; we can underestimate AI’s impact on this competition.

Amodei is wrong to see China as the only villain here. Clearly, the US is the greater threat—it is the only one of the two main imperialist powers that is actively bombing another country. But China is not a neutral party, nor a socialist state, and certainly not an anti-imperialist one. It is actively oppressing people in Hong Kong, Xinjiang, and Tibet. It’s flexing its military power and regional control in the Asia-Pacific, particularly against the Philippines and Taiwan. It’s also using its economic might, much like the US has done to gain market share and influence across the globe.

AI is extremely important in today’s imperialist rivalry because it is a dual-use technology. It has both civilian and military applications. So, every power is keen to develop AI as a weapon against its rivals. For instance, Russia has used AI in its drones in its imperialist war on Ukraine. Ukraine has used AI to defend its sovereignty. Both the US and China have watched and supported opposite sides, learning from the war how to develop AI in their own weapons.

Of course, there has been some hype about the technology’s military applications, just as there has been in the business world. Israel has been using AI to attack the civilian population of Palestine during their genocide, bragging in self-incriminating ways about their ability to target people and kill them and their families in their homes. But they didn’t need this technology to carpet bomb Gaza, they just wanted to advertise its use and therefore drive up sales of their AI applications.

Applying this tech to targeted assassinations may be useful. But more important is how AI enables states and non-state actors to wage cyber warfare on critical infrastructure, run autonomous machinery (negating Brecht’s assertion that tanks’ one weakness is that they require a driver), and analyze battlefield data in real time.

All the imperialist powers realize that AI is crucial in the competition for military supremacy. This is truly frightening in a world in which nuclear weapons exist, bio-weapons are still being developed, economic competition is increasing, and imperial rivalries are heating up. It does not bode well.

As great powers race to add AI to their stockpile of weapons of mass destruction, lesser powers will do the same. States like Iran, as well as non-state actors, use open-source models that are not far behind the frontier companies, develop their own AI weapons, and use them to avenge the egregious attacks on their populations by the US and Israel.

Anthropic’s claim that the Houthis have already been attempting to do this is but one example. No matter one’s motives, if you are involved in a military struggle, you have to develop AI as a weapon in your arsenal. The all-too-common analysis on the Left that AI is just hype will not help us understand the real world, where AI has already become part of military conflict with deadly results.

As far as the relative positions of the US versus China in this race, I don’t know enough to say. Anthropic’s plea to maintain the ban on exporting advanced chips to China may keep them behind, but it is unclear to me how successful that will be in the short or long term. It will put competitive pressure on China to develop its own chip-making technology and more efficient AI training methodologies.

The claims about China using distillation (essentially copying frontier models’ answers to train its own models) may be real, but as DeepSeek proved, it can also innovate on its own. It is hard to tell what capabilities the Chinese state has in this regard. Who would bet against it? A Sputnik moment is at least plausible.

So is an imperial conflict over Taiwan, which is the primary producer of the high-end microchips that power AI. China threatens to annex the country, which it regards as a renegade province, while the US wants to maintain its current status as its unofficial vassal.

These geopolitical imperialist rivalries can make us feel insignificant and powerless. But I think it is important to remember the politics of socialism from below as we consider them. These states get their power from their respective working classes. We have more in common with each other than with our bosses and state representatives. Nowhere is this clearer than the threat AI poses to workers on the job, from layoffs to speedups.

“Neither Washington nor Beijing but International Socialism” will be an important slogan to counter the nationalist jingoism espoused by the likes of Amodei. It can guide our actions as we build international bonds through organizing, in solidarity with our struggles and against our own oppressive states.

This interview will conclude in part three, proposing organizing interventions and strategies of resistance for the Left.

Opinions expressed in signed articles do not necessarily represent the views of the editors or the Tempest Collective. For more information, see “About Tempest Collective.”
Featured Image credit: David S. Soriano; modified by Tempest.