Harvard Built It. DARPA Paid For It. Nobody Governs It.

Scientists at Harvard’s Wyss Institute for Biologically Inspired Engineering have created something that did not exist six weeks ago: a tiny living robot with a functional nervous system that it built itself. No plug. No battery. No remote control. The little creature swims, explores its environment, and responds to drugs the way a nervous system is supposed to respond—because it has one. They call it a neurobot. To understand what that means, a bit of context is necessary, because this creature has been decades in the making.

It started in 2020, when the same Wyss Institute team created xenobots—tiny spherical structures assembled from the embryonic skin cells of Xenopus laevis, the African clawed frog, a species that has been a laboratory workhorse for decades. Cut a small piece of tissue from a frog embryo, drop it in a dish, and something strange happens. The cells don’t die. They heal themselves into a sphere, sprout hair-like projections called cilia across their surface, and start moving through water—with no scaffold, no genetic manipulation, and no instructions from anyone. Just cells doing what cells apparently do when removed from the body they were meant to build and then are left alone.

Xenobots could navigate, repair injuries, and collect loose cells to assemble copies of themselves. Remarkable, but their behavior was essentially mechanical—simple propulsion, basic environmental reactions, no internal control. The Wyss team wanted to know what would happen if they added one.

What They Did and How
Building a neurobot is, as one account put it, “a fiddly bit of microsurgery.” Cells from roughly 50 frog embryos are separated and kept apart for about three hours—long enough to commit to becoming neurons rather than skin. Those neural precursor cells are formed into small clumps. Meanwhile, a fresh piece of frog tissue heals itself into a sphere over about 30 minutes. During that window, researchers tuck the neural clumps inside before the tissue seals shut. Within two days, the composite heals. By day three, cilia appear on the surface, and it begins to move.

The implanted neural precursor cells differentiated into mature neurons on their own, extended axons and dendrites, formed connections with each other and with cells at the surface, and wired themselves into a functional nervous system—in a body plan that evolution never produced and nobody designed. It happened without instruction.

And guess what? The neurobots run entirely on their own biology. Multiciliated cells on their outer surface beat in coordinated waves, acting as a built-in biological thruster—no motors, no electronics, no external power. They survive for days on nutrients in their aqueous environment without any supplemental feeding.

Compared to earlier xenobots, neurobots move differently—they are more active, less predictable, tracing looping and spiraling paths. Researchers exposed them to a neuroactive drug called PTZ and they responded differently than non-neural biobots did. The neurons weren’t just present. They were shaping behavior.

Then things got stranger. Analysis of gene expression revealed unexpected activity in genes tied to visual system development—opsins, retinal cell markers—switched on despite the absence of anything resembling an eye. What those genes are producing, and whether they influence behavior, the team is still investigating.

The Human Cell Problem
The Wyss team’s next step is to replace the frog neural cells with human neural cells. A working platform for this already exists—”anthrobots,” built from clusters of human lung cells, have been moving around in labs for some time. The plan is to apply the neurobot framework to a fully human cellular context. A living machine, built from human cells, with a self-organized nervous system, that moves, responds to stimuli, and through “further conditioning and guided learning” becomes capable of adapting its behavior in programmable ways. What could possibly go wrong?

Michael Levin of Tufts University and the Wyss Institute, who led the research, does not shy away from the philosophical weight of that. “Where does form and function come from in the first place?” he asks. “When it’s not evolved and it’s not engineered, where do these patterns come from?” The nervous system inside the neurobot was not designed. It organized itself using rules that apparently operate even in a body that evolution never tested. Nobody fully understands why. Back in 2020, when the xenobot work first went public, Levin said something that applies with considerably more force today: “When we start to mess around with complex systems that we don’t understand, we’re going to get unintended consequences.”

Who Is Funding This?
The Wyss Institute was founded with a $125 million gift from Swiss billionaire Hansjörg Wyss—later doubled to $250 million—and has a long, well-documented funding relationship with the Defense Advanced Research Projects Agency (DARPA). DARPA (the covert group behind the rapidly advancing march towards transhumanism) contracts with The Wyss Institute include developing organs-on-chips worth up to $37 million, a $3.7 million contract to build a genetic security system for tracking the history of organisms, military exosuits, blood-cleansing sepsis devices, and suspended animation research. The original xenobot work—the direct predecessor to the neurobot—was funded in part by DARPA’s Lifelong Learning Machines program.

Haleh Fotowat, the first author of the neurobot paper—the scientist who actually built them—is currently the lead neuroscientist on a separate DARPA-funded program at the Wyss focused on anesthesia and biostasis: chemically inducing suspended animation, originally conceived to keep wounded soldiers alive in combat.

The profitable commercial angle was disclosed in the neurobot paper itself and largely ignored by everyone who covered it. Levin is a scientific co-founder and paid consultant for Fauna Systems, a startup built to commercialize frog cell-based biobot technology. He and co-founder Josh Bongard received $8.3 million in federal funding to develop scalable xenobot manufacturing. Levin also co-founded Morphoceuticals to commercialize his limb regeneration research. In other words, the man publishing academic papers on self-organizing living machines with nervous systems is simultaneously building the commercial pipeline to bring them to market.

And the agency most reliably funding the underlying science has made no secret of where it is headed. DARPA runs an active program called HyBRIDS—Hybridizing Biology and Robotics through Integration for Deployable Systems—explicitly combining biological and synthetic components for national security applications. Its Biological Technologies Office lists “forward engineering of biological systems”—cells, tissues, organs, organisms—as a core mission, alongside human-machine interfaces and neurotechnology. DARPA has, as one scholar documented, been “working on changing what it means to be human” for decades. The neurobot—self-organizing, self-powered, neurally active, and headed toward human cells—fits that trajectory.

The neurobot paper does not list DARPA as a funder of this specific study, but, with the documented connection, it does not need to.

No Rules Apply Here
There is no binding regulatory framework for living machines. None. Researchers publishing in the field on biohybrid robotics have called an ethics code of conduct for the field “an imperative task”—which is another way of saying that it does not exist. Bioethicists examining xenobots flagged years ago that these constructs are neither traditional robot nor known species—something genuinely new for which existing regulatory structures were not built.

That was before the nervous system. Before the human cells. Before the conditioned learning. Before the commercial scale-up. Before genes associated with visual perception started switching on in creatures that have no eyes.

Levin and Fotowat are asking real questions and doing what most would call real science. The problem is not their intentions. It is that the research is running well ahead of any public conversation about where it goes, who controls it, and what happens when a self-organizing living machine built from human neural cells does something nobody anticipated—in a body plan that evolution never tested, inside a commercial and defense ecosystem with its own very clear ideas about what comes next. Make no mistake, those who followed the mRNA platform from DARPA’s lab bench to a catastrophic global rollout without adequate public deliberation will find the pattern here quite familiar. And frightening.