Elon Musk recently brought back one of his long-standing ideas during the OpenAI trial: that as artificial intelligence becomes more powerful, humans may need a closer, more direct connection to machines.
He calls this future “human–AI symbiosis.” The phrase sounds ambitious, but the underlying idea is straightforward. If AI begins to outpace human cognition in both speed and capability, then interacting with it through keyboards, screens, and voice may eventually feel insufficient. Neuralink is Musk’s current answer to that problem: an implanted brain–computer interface that allows patients with severe paralysis to control external devices using neural signals.
The interface bottleneck
But the real ambition goes far beyond medicine. It is about redefining how humans relate to intelligence itself. That ambition deserves a far more serious and nuanced public conversation than it is currently getting.
“Safe brain-computer interfaces would be the culmination of an ongoing merge between humans and our intelligent machines that has been unfolding since the first programmable computer in 1939. From high-level programming languages, to GUIs, to vibe coding, the bandwidth of our communication with machine intelligence has steadily increased. Yet all these have been limited by the fact that we can think much faster and more richly than we can read, speak, or type,” comments John-Clark Levin, Research Lead at Kurzweil Technologies.
Today, when most people hear “brain–computer interface,” they think of medical recovery. A person who has lost the ability to move or speak regains a channel to communicate. That is one of the most powerful and meaningful applications of technology we have. It is clear, it is urgent, and it is deeply human.
But there is a second narrative quietly forming alongside it.
In that narrative, the brain is no longer only a site for medical intervention. It is being reframed as the next interface layer — the mechanism through which humans might keep pace with increasingly powerful AI systems.
That shift is not subtle. And it should make us pause.
A medical device designed for a patient with severe disability belongs in one category of discussion. A brain interface proposed as a general-purpose gateway into the AI future belongs in another entirely. The moment you move from one to the other, the questions change. Who gets access? What level of risk becomes acceptable? Who controls the infrastructure? And what happens when a medical technology starts evolving into a universal computing platform?
Surgical assumption
These questions become even more uncomfortable when the dominant path forward involves surgery.
Let’s consider the broader scientific goals behind BCI.
Levin says, “Curing diseases like Alzheimer’s and schizophrenia will require a richer understanding of the whole living brain than is possible with existing platforms. But the skull imposes physics-based limits on what can be recorded externally, and inserting electrodes with surgical approaches is inherently too destructive to safely scale to the needed bandwidth to gather that data. We need approaches that can unlock insights deep within the brain yet be introduced without surgery.”
For certain patients, brain surgery is a rational and justified choice. If someone cannot move or communicate, the potential upside of regaining that ability can outweigh risks that most people would never consider. That is the nature of medicine. The severity of the condition shifts the entire risk calculus.
But that logic does not translate to the general population.
A technology that is acceptable in extreme medical contexts does not automatically become a viable foundation for everyday human–AI interaction. The fact that AI is advancing quickly does not make invasive procedures broadly acceptable.
And yet, much of the public conversation skips over this gap entirely.
We move too quickly from extraordinary medical breakthroughs to sweeping claims about the future of humanity, without stopping to examine what it would actually take for an ordinary person to adopt a brain interface. Not in theory, but in practice.
If human–AI symbiosis is ever going to extend beyond a small group of patients, it cannot rely on a model that requires surgical intervention as the entry point. It cannot depend on hospitals, operating rooms, and specialized neurosurgical teams as a default pathway. That may be the right approach for some patients. It is not a scalable model for society.
The point is not to diminish the importance of implanted brain–computer interfaces. They are critical. In many cases, they will produce the strongest clinical outcomes. They will continue to play a central role in treating severe neurological conditions and restoring lost function.
The mistake is treating that path as the inevitable starting point for everything that comes next.
Early technological decisions tend to crystallize into long-term frameworks. If brain–computer interfaces are initially defined as invasive, expensive, and accessible only through specialized medical infrastructure, that framing does not stay contained. It shapes how investors think, how regulators respond, and how the public perceives the entire category.
And once that perception is established, it becomes very difficult to change.
That is a fragile foundation for a technology that is increasingly being positioned as central to the future of human interaction with AI.
A more important question is not how quickly we can connect brains to machines. It is what kind of connection would actually meet the threshold of real-world adoption. What would be safe enough, practical enough, and accessible enough to matter at scale?
There are researchers already exploring alternatives. Approaches that aim to interact with the brain without permanent implants. These ideas are earlier, less mature, and technically more challenging. They do not yet offer the same level of signal fidelity or control as implanted systems.
But they are asking the right question.
If brain–computer interfaces are ever going to move beyond a narrow clinical population, can they begin somewhere other than the operating room?
At this point, this is no longer just a technical discussion for companies and researchers.
The debate on infrastructure
As AI systems become more powerful, the interface layer becomes more important. If access to that interface is controlled by a small number of companies, delivered through a limited number of institutions, and restricted to a small group of users, then the future of human–AI interaction will be constrained from the outset.
What’s the future relationship between humans and AI? Two paths will emerge.
“Superintelligence can either become a separate species and likely escape human control, or it can be shaped into a tool that augments and amplifies humans,” says Levin. “Yet computers operate millions of times faster than the brain, so fully harnessing superintelligence with a chatbot interface won’t be possible. Safe brain-computer interfaces offer a pathway to keeping such systems aligned with human flourishing.”
We risk creating the most intimate computing platform in history — and simultaneously one of the least accessible.
It is tempting to assume that early adopters will be patients, and that broader societal implications can be addressed later. But that assumption is misleading. The core design decisions are already being made. The underlying architectures are being defined. The data, the standards, and the narratives are being established in real time.
By the time a technology reaches wider adoption, many of its foundational assumptions are already locked in.
Musk is right about one thing: the interface between humans and machines will become increasingly important as AI evolves.
But human–AI symbiosis should not begin as an elite medical procedure. It should not require society to accept brain surgery as the default gateway into the AI era. If the brain is becoming part of the technological frontier, then access, safety, and scalability cannot be treated as secondary concerns.
They have to be the starting point.
—
Tetiana Aleksandrova is a neurotechnology entrepreneur who has founded multiple companies in brain-computer interfaces. She is currently founder and CEO of Subsense.
