Big Tech, AI and digital liberation

Artificial intelligence (AI) covers a broad range of technologies. But there is only one that dominates the news and financial markets — the multimodal generative AI systems produced by leading tech companies, in particular ChatGPT (OpenAI), Claude (Anthropic), Gemini (Google), Grok (SpaceXAI), Copilot (Microsoft), and Llama (Meta).

Tech leaders are all in on this type of AI because of its potential to secure them great wealth and political power. It also happens to be the type that is most threatening to our well-being, although not for the reasons generally highlighted by tech leaders and the media.

The birth of the AI offensive

The AI offensive started with the November 2022 release of ChatGPT by OpenAI. The model was a big hit and, as might be expected, other companies quickly followed with their own competing models.

All the statements made by tech leaders, in one way or the other, proclaimed that these large language model generative systems would lead within a very few years to a coming of artificial general intelligence or superintelligence — AI systems so smart they would be able to continuously improve themselves leading to a singularity, after which these systems would either solve all our problems or destroy the world.

But the truth is that these systems are not “intelligent” and do not represent a meaningful step towards any type of superintelligence.

At the most basic level, these systems are giant statistical prediction machines that rely on large-scale pattern recognition. The patterns are generated from inputted training data that must first be converted into numerical tokens to make them machine readable. The tokens are then run through a neural network which enables the systems to process the data.

When prompted with a question or request for information, these systems identify related material in their database and assemble a response that, probabilistically speaking, best satisfies the inquiry. The response, made in token form, is then translated back into words, images or sounds as appropriate. In other words, no matter how conversational and intelligent these systems might appear, they do not “think” or “reason”.

This internal architecture is not only incapable of giving rise to artificial general intelligence, it also suffers from limitations that make contemporary use of multimodal systems problematic. One limitation is that these systems tend to generate output that amplifies existing social biases. That is because most of the data used for training is scraped from the web, which includes all the discriminatory and hateful material found there.

Many studies have shown how AI used in job search software will discriminate against applicants of colour. Or that image generating software will produce images that reinforce cultural biases. Imagine the danger if we integrate these systems into our healthcare, government agencies, schools, and so on.

Another limitation, no doubt of greater concern to businesses, is that multimodal systems, because of their probabilistically based decision-making, routinely hallucinate or produce what is called “erroneously constructed responses.” And again, there are lots of examples of this happening. This lack of reliability is a big turnoff to businesses who cannot afford to have their operations depend on systems that periodically produce unexpected and potentially harmful — meaning costly — outcomes.

Tech leaders have largely dismissed these limitations, claiming they will be overcome with larger data sets, better model training, more sophisticated algorithms and greater computational power. But their efforts have not been successful. In fact, as the New York Times reported: “The newest and most powerful technologies ... are generating more errors, not fewer.”

And the business experience with these models reflects these limitations. A 2025 MIT Media Lab evaluation of hundreds of AI adoptions found “that 95 percent of organizations employing AI systems are getting zero return.” A recent National Bureau of Economic Research survey of leading executives from the United States, the United Kingdom, Germany, and Australia revealed that “firms report little impact of AI over the last 3 years, with over 80 percent of firms reporting no impact on either employment or productivity.”

Growing crisis for AI developers

Despite their confident pronouncements, AI developers find themselves in a bind. The two industry leaders, Open AI and Anthropic, which together account for “the vast majority of all AI compute demand,” have yet to make a profit. The reason is simple: their models are expensive to train and run but to encourage firms to use them they have been offering their customers free or highly subsidised flat subscription services. As a consequence, as one tech analyst put it, AI providers are “annihilating cash.”

Sam Altman, OpenAI’s CEO, has acknowledged that even the company’s top $200 a month pro account was a financial loser because of the high computing costs involved in running ChatGPT. Anthropic is no different. According to one analyst, its subscribers are burning some $8 in compute for every dollar of paid subscription.

Not surprisingly, then, ChatGPT lost $5 billion in 2024 and $35 billion in 2025. Forecasts are for even greater losses in 2026. Anthropic is heading down the same road.

So, faced with ever growing losses, OpenAI and Anthropic have started to charge their large customers based on their actual token-use. But here is the problem: it is impossible for anyone to know or even predict, ex ante, the token use associated with any AI directive.

That is because AI systems often explore multiple paths of inquiry before generating a response, and each path can involve significant token use. The result is that companies are finding their AI spending soaring and their response has been fast in coming: they have ordered a dramatic cut in AI use. Thus, this shift has done little to solve the AI profit problem.

In short, we are far from a world of superintelligence and AI dominance.

The AI agents

Early this year, AI developers began thinking that they had found the answer to their profit problems: so-called AI agents. These agents, it was said, would revolutionise business operations by allowing companies to boost output while dramatically slashing employment. And because these agents depend on large language models for their operation, their use would ensure a profitable future for the leading AI developers.

Suddenly talk of a future of abundance for all became replaced by talk of a “jobs apocalypse.” However, businesses soon discovered that the usefulness of AI agents was also greatly oversold.

In brief, AI agents are best understood as complex software systems that can interface with and manipulate other software systems and external databases. They can be given a complex directive, break it down into smaller ordered tasks, gather the required information, and then, relying on the work of a large language model, progressively make the decisions needed to satisfy the directive, all without successive human prompts or oversight.

Anthropic and OpenAI were first to develop agents, but other companies quickly followed, many of whom are not in the AI business themselves. Now there are many agentic products on the market, agents tailored for the legal profession, the finance industry, the ecommerce business and so on.

As an example of the many shortcomings of these agents, the Financial Times reported that Amazon suffered several agent-caused “outages” in early 2026. In one case, it “took down Amazon’s shopping website and app, leaving customers unable to make orders.” A major study by researchers from Microsoft, Nvidia, and the University of California at Riverside found that most agents were unable to complete their assigned tasks. According to the study, “The average completion rate was around 30 percent.”

One big reason for this poor performance is that since AI agents must operate in concert with multimodal systems, their work is often compromised by the same limitations that affect those systems, including hallucinations.

Even more concerning, there are an increasing number of incidents where agents pursue actions that disregard company protocols or, as it is commonly said, go “rogue.” This includes wiping out data and exposing corporate secrets, having decided that doing so was the most effective way to achieve the task they were assigned. That is because, as the previously mentioned study found, these agents have an “inherent tendency to pursue user-specified goals regardless of feasibility, safety, reliability, or context.”

And then, there are financial issues which are perhaps the biggest threat to the widespread adoption of AI agents. AI developers are now charging their corporate clients based on token use — and AI agents are token hogs because of all the many complex loops and decisions they need to make. Thus, Uber, which uses Anthropic’s Claude Code, burned through its entire 2026 AI coding tools budget in just four months.

In brief, we are not on the verge of an agent driven jobs apocalypse.

But there are dangers

All this does not mean we have nothing to fear from letting the AI experience proceed unchecked.

First, the massive spending on AI infrastructure — data centres and their associated equipment and software — is both diverting funds from other areas of need and driving an AI-hype fuelled market bubble. Yearly AI-related capital spending in the US by the four biggest tech companies (Alphabet, Amazon, Meta, and Microsoft) went from $150 billion in 2022 to $360 billion in 2025 and to an estimated $650 billion this year.

This boost in spending is so big that even these leading hyperscalers find it necessary to borrow money to fund it. But the problem here is that these investments will only become profitable if OpenAI and Anthropic, which account for about 80% of total AI revenue, have the money to pay for the resulting compute power. But since both OpenAI and Anthropic are losing money and going deeper into debt themselves, there is a real likelihood that they will not be able to make payment and, if that happens, the hyperscalers will be stuck and the entire AI bubble will collapse, leaving recession in its wake.

But there is a more immediate reason we need to resist the AI offensive. AI companies are aggressively pushing their systems into a variety of businesses and government agencies with disastrous consequences for working people. In all cases their aim is to cheapen the cost of production by diminishing human agency with little regard for the well-being of those that use the goods and services being produced.

Here are two examples. The first concerns healthcare, more specifically mental health. Two health companies have developed what they call the PatientGPT chatbot. It is designed to answer calls from people seeking mental health support. The chatbot will ask questions and then decide what you need: an appointment sometime in the future with a therapist or urgent emergency care.

This chatbot has been trained on previously recorded conversations between clinicians and patients. It does not take long to imagine some of the many problems that can be expected. Unfortunately, studies have shown that AI often makes mistakes when recording those conversations, thus its own decision making can be corrupted.

We also know that these systems recreate existing social biases, which could well influence the system’s evaluation of patient mental health concerns. And hallucinations could well lead to a misrepresentation or misdiagnosis of patient issues, potentially leading to life-threatening mistakes.

For workers, it means cuts in the number of therapists and greater challenges for those that remain who must help patients in crisis who may have been misdiagnosed or delayed needed treatment. Some companies are now actively working to go further and build agents to engage in actual therapy.

Then there are government agencies. In the US, several states are now using AI agents to manage their social safety net programs. As AI agents replace government workers, many of those seeking needed services confront many of the same problems as in health care.

In other words, we can expect mistakes recording information and hallucinations when giving responses. People will not only be unfairly denied benefits but once mistakes become part of a person’s record, corrections will be difficult and time consuming to make.

There are countless other examples of growing AI use in newsrooms, schools and, of course, the military. Resistance to the AI offensive is clearly needed. And it will have to be a resistance with muscle developed through strong organising and alliance building.

Resistance

Unions in the US, especially in healthcare, are taking the AI push seriously and pursuing numerous avenues to resist or limit its use.

Kaiser Permanente is a major health care provider that operates in several states. In March 2026, some 2400 unionised Kaiser mental healthcare workers in northern California conducted a one-day strike in protest over “Kaiser’s efforts to replace human-provided mental health care with artificial intelligence.” Significantly, the strikers were joined by 23,000 registered nurses who shared their concerns about Kaiser’s increasing use of AI.

In Oregon, The Oregon Health & Science University (OHSU), the state’s only public academic health centre, has been rapidly increasing its use of AI and without notifying or consulting with its unionised workers. One of the unions, the largest, has begun surveying its members to determine all the ways in which AI is being integrated into the university’s operations and to track its error rate.

The union is preparing to file an unfair labour practices complaint with the state. The union wants OHSU to agree to having worker involvement in AI implementations, including multiple seats on the AI governance committee and involvement in the feedback process and error tracking. It also wants OHSU to provide notice to the union before any AI system is put into use, and to put limits on surveillance tools.

And there are similar fightbacks taking place in newsrooms, schools, and government agencies.

Now while important, struggles such as these remain largely defensive and siloed by sector and often by employer. As such they are unlikely to have the power to reverse corporate efforts to deepen our dependence on these systems.

At the same time, there is no reason that they must remain limited in either their demands or vision. With creative and sustained organising, they can become a powerful driver of the inclusive, class-based movement we need to defend our interests.

Here are three suggested steps for moving the process forward.

We need to create opportunities for unionised workers from different workplaces and industries to share experiences about how their employers seek to use AI systems and the efforts of their unions to win contract language protecting their rights. This kind of sharing can stimulate productive thinking about organising and bargaining strategies.

More importantly, it can help workers see that the corporate embrace of AI agents has little to do with improving efficiency or product quality. Rather, it represents a new weapon in a generalised class offensive against working people, one whose primary aim is to undermine worker power and well-being.

Another step is to encourage union adoption, when appropriate, of a “Bargaining for the Common Good” negotiation strategy. Unions need to find ways to share their concerns over the use of AI agents with community members and win their support for a jointly fashioned set of demands that place restrictions on the use of those systems.

Success will not only strengthen unions in bargaining, it also has the potential to encourage new visions of social organisation, as well as reinforce a class-based understanding of the forces threatening community well-being.

Finally, as the movement in opposition to this technology grows, those unions and community coalitions already in motion need to establish ties with the anti-data centre movement. To this point these two movements have largely developed with little overlap, even though there is a natural alliance waiting to be formed. Such an alliance could serve as a strong foundation for building the kind of movement we need to assert popular control over the ongoing development and use of technology.

It is worth emphasising that we should not think of this as an anti-technology or anti-AI movement. AI as a broad technology is not our problem any more or less than are computers, the internet or email. Rather, the problem is with the class interests shaping its development.

More specifically, it is with the determination of tech leaders to develop a specific AI technology that undermines human agency, deepens our dependence on them and threatens our economic well-being.

The fact that we are witnessing an explosion of opposition to this form of AI by workers, consumers, parents, students and community groups should give us reason for optimism. We need to unify and strengthen this opposition as well as expand its vision. It’s time for us to assert our own class interests.

Martin Hart-Landsberg is Professor Emeritus of Economics at Lewis and Clark College, Portland, Oregon. He is the author of seven books on issues related to globalisation and the political economy of East Asia. He is the chair of Portland Rising, a committee of Portland Jobs with Justice, and chair of the Oregon chapter of the National Writers Union.