Happy anniversary, Alexander Graham Bell. What have we learned in 150 years?

March 10, 2026, marks the 150th anniversary of Alexander Graham Bell’s first telephone call. His words—“Mr. Watson, come here; I want to see you.”—were simple but launched a revolution in human connectivity.

Ninety-three years later, another revolution began with a far less dramatic message. On Oct. 29, 1969, researchers attempting the first ARPANET connection between UCLA and Stanford tried to type the word “login.” The system crashed after the first two letters, so the first message ever sent across what would become the internet was simply “Lo.”

Those two moments—Bell’s call and ARPANET’s stutter—bookend a critical lesson for the AI era. Each communications revolution builds on the last, but each creates its own economic logic and governance challenges. Policy mistakes happen when we assume the future will behave like the past.

The telephone network evolved into what economists call a “natural monopoly.” The logic was that duplicating wires and switches made little sense, so each area gravitated toward a single network. But in exchange came obligations such as requiring nondiscriminatory access and interconnection.

That principle of open access mattered more than anyone anticipated. When government researchers began linking computers in the late 1960s, they did not build an entirely separate infrastructure but rode atop the existing phone network. Nondiscriminatory access to the telephone network helped ARPANET move beyond a research experiment and ultimately scale to become the internet.

The internet quickly diverged from the telephone model. While the underlying connectivity remained a for-a-fee telecommunications service, innovators built advertising-supported services on top of it. The broad term “internet” began to blur the distinction between the pathway and the platforms built on it. That shift did not diminish the importance of networks but shifted economic value toward the applications built on top of the network.

Many of the companies that rose to dominance through platform applications enjoying open access to the internet resisted regulation. They argued it would smother innovation. For a time, that argument carried the day; policymakers embraced a hands-off ethos, reinforced by the belief that digital markets were inherently self-correcting.

But the internet’s history is more complicated than such mythology suggests. It was not born in a regulatory vacuum. Government funding paid for development of the early internet. Antitrust actions—such as the 1956 AT&T consent decree that opened a previously closed patent portfolio, the 1969 unbundling of hardware and software by IBM, and nondiscriminatory network access—opened space for applications competition. Open standards bodies ensured interoperability. The internet’s dynamism flowed not from the absence of rules but from structural decisions that preserved openness at critical layers and enabled innovation in applications.

Over time, however, new chokepoints emerged. The same openness that enabled startup innovation also allowed their concentration. Applications such as search, social media, mobile operating systems, and digital advertising each became dominated by a handful of platforms. Competition gave way to gatekeepers acting to advantage themselves.

We now stand at a similar inflection point with artificial intelligence (AI).

Like the telephone and the internet network before it, AI is a general-purpose technology that enables other innovations. But unlike those earlier systems, AI is not just a communications medium. It is programmable cognition—a technology that generates decisions, content, and increasingly, actions. And like its predecessors, it is being shaped by structural choices made in real time.

Those choices increasingly center on control of the AI stack—the microchips, cloud computing, foundation models, developer tools, and distribution channels that are essential to the provision of innovative AI applications. Increasingly, these essential capabilities are controlled by a handful of companies whose power traces back to their access to the infrastructure of the early digital era.

This is where drawing the wrong historical lesson becomes dangerous. Some argue that because the early internet flourished under relatively light regulation, AI should follow the same path. But that interpretation confuses outcome with origin. The internet’s early openness was not an accident. It was the byproduct of policy decisions—some intentional, some serendipitous—that prevented infrastructure owners from sealing off new opportunities too early.

Today, many of the same firms that built their fortunes by benefitting from that openness are racing to control the capabilities necessary for AI applications—all while simultaneously building their own applications. They invoke the language of innovation and global competition to argue against oversight. But history points to a better conclusion. When a general-purpose technology becomes foundational infrastructure, the issue is not whether rules will emerge—the issue is what kind of rules will promote and protect the next wave of innovation.

The deeper lesson of Bell’s telephone and ARPANET’s “Lo” is not about nostalgia. It is about structure. Transformative technologies consistently follow a pattern: Early openness enables experimentation; consolidation creates bottlenecks; society eventually confronts the need to rebalance power to sustain innovation and public trust.

Leading AI model developers now appear to be entering the middle phase of consolidation to extend their control downstream. The outcome of such architectural decisions will echo for decades. Choices about interoperability, openness, and access will determine whether AI evolves like the early internet, with broad-based innovation, or like later digital platforms, where control concentrated behind proprietary walls.

The real question is not whether AI will be governed, as every transformative technology eventually is. The real question is whether we learn from the layered history of networks and act early enough to preserve openness, competition, and democratic accountability—or whether we once again wait until private power has already hardened into digital infrastructure.

From Bell’s call to Watson to ARPANET’s first flicker of connectivity, history has been generous with clues. The challenge now is whether we are wise enough to use them.