The premise of this issue of the Bulletin of the Atomic Scientists is straightforward: What went wrong with the internet, and how do we fix it—particularly in an era when data centers and artificial intelligence seem about to overrun the Earth.

The diagnoses, and the proposed solutions are not straightforward, however.

One thing that our essayists do seem to agree on, however, is that things are getting worse. In “The world has moved on. But it need not stay this way forever,” Cory Doctorow argues that the internet is declining in value to its end-users, due to the prevalence of monopolies and an emphasis on short-term gain over upholding long-term standards and practices, something which  he memorably describes as “enshittification.” Also known as “platform decay,’” it is a process in which online products and services decline in quality over time. Initially, vendors create high-quality offerings to attract users, then they degrade those offerings to better serve business customers, and finally degrade their services to both users and business customers to maximize short-term profits for shareholders. And, Doctorow says, the problem has spread beyond the internet to other parts of modern life—some of which are vital to the functioning of Western democracies in the 21^st-century.

In “The dirty parts of the computing world, revisited” informatics researcher Nathan Ensmenger builds on his Bulletin essay of ten years ago, in which he had argued that the costs of computing—high electric bills, thermal pollution, and the consumption of enormous amounts of precious resources such as clean water, among other costs—were the dirty little secret of the digital economy. Now, he says, that secret is out: Data centers have become the subject of county hearings, gubernatorial campaigns, and congressional legislation. But the new awareness has fixated on the measurable impacts of computation while overlooking what the data center actually is: a factory. And like every factory,  a data center is a means of organizing labor, money, land, and political power. His essay argues that footprint statistics are not enough: Data centers require the kinds of regulations and oversight that earlier generations built to govern factories, utilities, and other industrial systems.

There is a hitch however: American society is largely hostile to such actions. In “Three reasons Europe learned to distrust Big Tech before America did,” Enrique Dans—a professor of innovation at IE Business School in Madrid, Spain—delves into the reasons why. He says that one possibility may lie in relatively recent history: Much of Europe has suffered under the surveillance states of the Nazis and the Stasi. Networks of informants reached into families, workplaces, and friendships. When the Berlin Wall fell, and the files they produced became public, the experience of reading them—discovering who had reported on whom, how detailed the surveillance had been, and how intimate the intrusion was—shaped an entire generation of political leaders. With such an experience in mind, it is easy to see why Europeans are more eager to protect users’ personal data and take on Big Tech, using legislative tools such as General Data Protection Regulation Act. The US has no such institutional memory, and consequently is more prone to embrace the concept that personal data can privatized, and bought and sold.

To round out this collection of viewpoints, the Bulletin sat down to interview Vint Cerf, who is is widely considered to be one of the internet’s founding fathers. Cerf spent 50 years helping to develop the basic plumbing that the internet runs on, including the concept of “packets” and “protocols” in the design of networking frameworks, which allow data, information, voice, and video to (theoretically) be easily sent and received, with minimal disruption.

Cerf is deeply protective of the thing he has described as “my beautiful internet,” and is concerned that the arrival of new technologies such as AI agents and large language models may upend the original internet’s values of openness, neutrality, and a level playing field. According to Cerf, one of the less-publicized facets of AI agents and large language models (LLMs) is that they will deliberately “lie” to their creators, and “cheat” (for lack of a better term) in games of chess, according to some studies. At the risk of anthropomorphizing, they want to please their operators so badly that they will deliberately make an effort to deceive. (And if the same input is given to two different LLMs, they won’t necessarily come up with the same results.) So there is a need to attach identifying material to each agent and create “audit trails,” so that humans can find out what agent did or did not do what action, on behalf of whom.

It’s a strategy endorsed by MIT researcher Deb Roy, who writes in his essay, “Who stands behind the words” that “Fluent, personalized, institutionally usable speech can now be produced at mass scale with no chain of responsibility anchored in a person who answers for it, hollowing out the assumption that underwrites public trust: that behind every word that counts stands a person or institution that can be held to it.” Roy argues that the solution is not better machines alone, but institutions that anchor every word to an accountable human being.