I thought I had heard incorrectly the first time someone told me that human neurons floating in a shallow dish at a Melbourne lab had learned to play Pong in about five to twenty minutes. I hadn’t. In 2022, scientists at Cortical Labs cultivated living brain cells on an electrode grid, connected them to a condensed version of the classic Atari game, and observed how the cells learned to paddle. The icy, mechanical mastery of a trained neural network eluded them. They were not required to. They were acting strangely—adapting in the manner that living things do.
The experiment did not remain in a dish. The idea of combining biology and computation has always had a sort of religious appeal in Silicon Valley, where it quietly permeated the local culture. D. Scott Phoenix, wearing a black turtleneck and the required tiny ear mic, told the audience at TED 2026 in Vancouver last month that the merger is no longer speculative. “Someone you work with will get it first,” he replied. “And just like you did with the smartphone, you’ll wait. But you won’t in the end.” It’s difficult to ignore the fact that he is now a venture capitalist; it’s the kind of line that falls somewhere between prophecy and sales pitch.
Phoenix is not by itself. In 2017, Sam Altman proposed the merger, describing it as “probably our best-case scenario.” In the same way that others discuss their workout regimens, Peter Thiel promotes transhumanism. Additionally, the money is traveling with them. The market for brain-computer interfaces, which is currently valued at about $350 million, is expected to grow to $1.2 billion by 2035, and the neurotechnology industry as a whole is anticipated to soar to $52 billion by 2032. After being financed by Elon Musk and Mark Zuckerberg, Phoenix’s former business, Vicarious, sold to Alphabet in 2022. You are already familiar with the pattern.
The simplest case to defend is the medical one. After receiving a Neuralink implant, Noland Arbaugh—who was paralyzed from the neck down—became the first person to use his thoughts to control a computer. Something tightens in your chest when you watch that kind of footage. Giving people back pieces of their lost lives is a good idea. However, the medical frame is also becoming more and more of a warm-up act. Wearable headbands, smart glasses, stress-tracking apps, and other devices that gather signals from millions of healthy, functional brains are the true commercial prize.
Here, the original Stanford researchers—the same group that initially taught the Valley how to model attention, reward, and prediction—have begun to rebel. In an effort to keep neural data out of the typical data-broker economy, a number of them currently serve on boards or advisory groups. Former UNESCO chief Audrey Azoulay wrote in the Financial Times, “If data is the oil of the 21st century, then ‘brain’ data is the crude oil.” Although the metaphor is awkward, it effectively conveys the anxiety.
Legislators have subtly changed privacy laws in politically disparate states like Montana, Colorado, California, and Connecticut to include information produced by an individual’s “central or peripheral nervous system.”A more comprehensive framework for neurodata rights is being considered in Minnesota. Chuck Schumer, Maria Cantwell, and Ed Markey introduced the MIND Act in September of last year, which would require the FTC to investigate how neural data reveals emotion and thought. Rafael Yuste, a neuroscientist, founded the Neurorights Foundation in 2022, which is largely responsible for this drafting work.
Not everyone believes that changing outdated privacy laws is sufficient. The Duke researcher who has spent years studying this, Nita Farahany, contends that brain data should have its own category that is distinct, more stringent, and more difficult to commercialize. She once casually mentioned to me, “The most intimate data is the data about what you’re thinking and feeling.” Listening to her gives me the impression that statutes aren’t the main point of contention. It concerns whether a person’s final unmapped territory remains unmapped. Nobody really knows if that is still feasible.⁖※
