Tech • AI • Robotics • Game
Black Mirror in your brain — is it coming soon?
7/10TL;DR
A Palo Alto start-up’s $50 million funding round for a thought-reading cap has intensified debate over brain-computer interfaces, as advocates argue direct links between human brains and AI agents may become necessary to preserve human control over increasingly powerful artificial intelligence.
KEY POINTS
From proof of humanity to brain links
Saturnin Pugnet, an early member of the founding team behind World, frames the rise of brain-computer interfaces as part of the same response to advanced AI that motivated large-scale digital identity systems. World, co-founded by Sam Altman, was built around “proof of humanity,” using iris-based verification to distinguish real people from bots and deepfakes at internet scale.
A manifesto arguing for neural connection
In a 168-page manifesto titled Face Lock, Pugnet argues that keeping humans in control of AI may ultimately require connecting the brain directly to machines. His core claim is that the same scaling logic that improved AI models through more data and compute could also transform brain decoding, provided enough neural data can be collected and modeled.
Brain decoding is already further along than many assume
Pugnet says non-invasive systems can already extract meaningful information from neural activity without surgery or implanted chips. Using headsets and electrodes, researchers can reconstruct the semantic content of imagined text, images and even videos, not as exact pixels but as concepts such as an aircraft taking off or an object moving in a direction.
A new hardware race is emerging
The commercial push is accelerating. A Silicon Valley company recently raised $50 million for a cap designed to connect human thoughts to AI agents, while neurotechnology start-ups raised $4.8 billion in 2025, roughly double the previous year. In January, OpenAI joined a $250 million seed round for Merge Labs, and in March China authorized a brain implant presented in near-commercial terms.
Why the technology appeals to AI builders
The argument is not only about human augmentation. It is also about data. Pugnet contends that today’s models are trained on outputs of the human mind such as text, images, audio and video, but not on the underlying neural processes that produce them. He describes internet text as the “subtitles” of cognition rather than the original signal.
A possible new training dataset for AI
In that view, neural data could become a new frontier as the supply of public text for model training plateaus. Pugnet estimates that 13 million hours of brain recordings could rival a major modern training run, and that assembling a corpus of 10 million hours might cost around $3 billion, potentially cheaper than some current large-scale AI training efforts.
Consumer use cases go beyond medicine
Beyond restoring speech or movement, advocates envision direct thought-based interaction with software. The promise is to remove the compression imposed by keyboards, touchscreens and spoken prompts, allowing an AI system to grasp intention more precisely. That could speed up writing, image editing, design and other iterative tasks by translating mental feedback into immediate changes on screen.
Identity, accountability and ownership
Supporters also tie the technology to authentication in an AI-saturated internet. A neural link combined with systems like World could, in theory, help verify not only that a human exists but that a specific action was intentionally authorized by that human. Some even speculate about registering the first emergence of an idea and attaching ownership or royalties to it, potentially through blockchain-based records.
Privacy fears remain central
The strongest objections concern mental privacy, surveillance and dependency on private platforms. Critics fear a future in which access to cognition is monetized, manipulated or restricted, with intimate thought data treated more invasively than today’s smartphone or social media records. The concern is heightened by the belief that the brain remains the last truly private domain.
The geopolitical race is sharpening
China and the United States are seen as the most likely leaders because both can mobilize large capital pools and tolerate more radical experimentation than Europe. Advocates argue competitive pressure will drive adoption if early users gain major productivity advantages, while skeptics warn that social and ethical safeguards may lag far behind the technology.
An attempt to keep AI aligned with humans
Pugnet argues that neural interfaces could help solve AI alignment by exposing systems to richer traces of human intention, intuition and emotion. Without that deeper signal, he says, AI may remain highly capable but oddly alien, excelling in some domains while missing basic human context. The broader warning is that AI progress is moving so fast that institutions may struggle to absorb its consequences before the human-machine boundary is redrawn.
CONCLUSION
The emerging market for thought-to-AI devices is turning brain-computer interfaces from speculative research into a strategic technology race. The central question is no longer only whether the tools can work, but whether societies can govern access to the mind before AI makes such access economically and politically irresistible.
Ask a question