Generative AI is currently in the early stages of what some call enshittification. Today, it may still carry the scent of roses—but that won’t last. And don’t think you’ll simply turn your nose up at it when the rot sets in. That’s not how enshittification works: even when it reeks, we’ll keep coming back for more. Perhaps a definition helps—enshittification, or platform degradation, refers to the process by which an online service becomes steadily worse for users, as the provider shifts its priorities toward maximising profit.
Take Facebook as an example. In its early days, the platform was fully focused on its users—you’d see posts only from friends and family, with the occasional heart-warming reconnection with an old acquaintance. As it became harder to leave (since doing so meant losing all your social ties), Facebook pivoted to monetisation. Sponsored posts from advertisers began to appear between your childhood memories and holiday snaps. Gradually, the original experience was buried beneath a layer of digital detritus, while Facebook’s shareholders watched their coffers swell.
Monopolies
In the case of enshittification, products such as Facebook, Google, or TikTok are offered either free or at very low cost in order to hook as many users as possible. Thanks to billions in investment, tech firms could afford to run at a loss long enough to crush competition and establish monopolies. Only then did they switch to profit-making—at the user's expense.
Generative AI companies are in the early phase of this same trajectory. Basic versions of ChatGPT, Gemini and Claude are free, and even their premium offerings are priced far below their actual costs. OpenAI, for instance, posted a $5 billion loss in 2024. Revenue from paying users ($3 billion) covered only a fraction of the cost of training and operating their AI models. The company forecasts that it won’t turn a profit until 2029—by which time cumulative losses will have reached $44 billion.
Meanwhile, OpenAI and Google are offering their respective chatbots free of charge to American universities, hoping to win over hundreds of thousands of students. Microsoft’s Copilot is trying to make itself indispensable in as many workplaces as possible. These companies are chasing new monopolies. Unless they suddenly decide to start charging real costs, the next stage of enshittification is inevitable—and the user experience will change drastically. Advertising needn’t be overt; a chatbot can steer your preferences in far subtler ways. These bots already appear more addictive than social media, and their habit of telling users what they want to hear doesn’t bode well for individual psychology or societal cohesion. But advertisers and shareholders? They'll be thrilled.
The price of "free"
In Belgium, selling goods at a loss is banned, to prevent monopolistic practices. That restriction doesn’t apply to services—perhaps with good reason. But in the digital world, does that logic still hold? The monopolies that tech giants are building certainly don’t benefit the public. The EU’s Digital Markets Act of 2022 aims to address this, but whether it will make a meaningful difference remains to be seen.
The saying goes, there’s no such thing as a free lunch, and that holds true here as well. Someone always ends up paying the bill. Sometimes it's us, when we get hooked on social media, manipulated without realising, and suffer the mental consequences. Sometimes it’s others who bear the burden—like the tens of thousands of Kenyan workers paid less than $2 an hour to fine-tune ChatGPT, exposed to traumatic content involving child abuse, bestiality, murder, suicide, torture, self-harm and incest in the process of censoring the AI’s raw output. Or we all pay together. These power-hungry AI models are sending tech firms’ energy consumption soaring, with climate pledges quietly shelved. Microsoft now emits 29% more greenhouse gases than five years ago; Alphabet (Google’s parent company) is up by 50%. In both cases, AI is largely to blame.
Paying the true price
We’ve grown used to internet services being free. But the psychological, ethical, political and environmental costs they incur are no longer defensible. Sooner or later, they’ll come back to haunt us. That’s why we need policymakers and business leaders to start charging a fair price for digital products—one that accounts for climate impact, decent wages, development costs, and maintenance. Only then will these products become sustainable, avoiding some of the more damaging side effects: privacy breaches, manipulative and polarising algorithms, exploitative labour practices, and ecological disaster.
And we, as users, must understand that nothing comes for free. Our current consumption of digital services simply isn’t sustainable. AI and other technologies can do a great deal of good—but only if we’re willing to pay their true cost. This may not be a message people want to hear. But perhaps our very idea of what makes us happy is part of the problem.
"This is a machine translation. We apologise for any inaccuracies."