The Mind-Machine Interface Is Here: Researchers at Columbia University, Stanford, and Penn have unveiled BISC—a brain-computer interface thinner than a human hair that wirelessly streams neural signals to AI at 100 Mbps, enabling breakthrough treatments for paralysis, epilepsy, and blindness while raising profound questions about mental privacy.

Brain Chips That Read Your Thoughts: The BISC Breakthrough

Discover BISC, the ultra-thin brain implant creating a wireless highway between your brain and AI. Learn how it decodes thoughts, its medical promise for treating neurological conditions, and the privacy questions it raises.

Reading time: ~14 min • Updated: Jan 15, 2026

Key Facts (TL;DR)

Ultra-thin design: BISC is just 50 micrometers thick—about the width of a human hair—and slides between skull and brain like wet tissue paper
Massive electrode count: 65,536 electrodes on a single silicon chip with 1,024 recording channels and 16,384 stimulation channels
100× faster than competitors: Wireless data transmission reaches 100 Mbps—at least 100 times higher throughput than any competing wireless BCI
1/1000th the size: Entire implant occupies approximately 3 mm³—less than one-thousandth the volume of conventional BCIs
AI-ready architecture: High-bandwidth recording enables advanced machine learning to decode movement, perception, and intent in real time
Clinical trials underway: NIH-funded epilepsy trials beginning; short-term human recordings already completed successfully
Privacy concerns unresolved: No regulations yet govern who owns neural data or how it can be used

Imagine a computer chip so thin it handles like wet tissue paper, yet powerful enough to capture 65,536 channels of brain activity and transmit that data wirelessly at 100 megabits per second. That's BISC—the Biological Interface System to Cortex—a brain-computer interface that represents a fundamental leap in how humans connect with machines.

Published in Nature Electronics in December 2025, BISC condenses an entire brain-computer interface onto a single silicon chip smaller than a postage stamp. Unlike bulky predecessors requiring skull-mounted canisters or chest implants, BISC slides into the narrow space between your skull and brain, opening new possibilities for treating epilepsy, paralysis, ALS, stroke, and blindness.

But BISC's promise extends beyond medicine. By creating a high-bandwidth link between the brain and artificial intelligence, it raises profound questions: What happens when AI can decode your intentions before you act on them? Who owns the data streaming from your neurons? And are we ready for a world where thoughts become as accessible as text messages?

What Is BISC and Why It Matters

The brain-computer interface revolution

Brain-computer interfaces aren't new. Researchers have been recording neural signals for decades. What makes BISC revolutionary is integration—cramming everything onto one chip rather than relying on external racks, skull canisters, or wired connections.

Developed by Columbia University, Stanford, University of Pennsylvania, and NewYork-Presbyterian Hospital, BISC represents a complete rethinking of BCI architecture. Lead engineer Ken Shepard explains: "Most implantable systems are built around a canister of electronics that occupies enormous volumes of space inside the body. Our implant is a single integrated circuit chip that is so thin that it can slide into the space between the brain and the skull, resting on the brain like a piece of wet tissue paper."

What problem does BISC solve?

Traditional BCIs face three core limitations:

Size: Conventional devices require bulky titanium canisters implanted in the skull or chest, connected to the brain via long wires
Bandwidth: Most wireless BCIs transmit data at 1 Mbps or less—too slow for complex AI decoding
Invasiveness: Penetrating electrodes that pierce brain tissue cause inflammation and signal degradation over time

BISC solves all three. Its subdural placement (between skull and brain surface) avoids tissue damage. Its single-chip design eliminates bulky external hardware. And its 100 Mbps wireless link provides bandwidth comparable to home internet—fast enough for AI to decode complex neural patterns in real time.

How BISC Works: From Brain Signals to Digital Data

The chip itself: 50 micrometers of silicon

BISC is a single CMOS (complementary metal-oxide-semiconductor) integrated circuit thinned to just 50 micrometers—roughly the diameter of a human hair. Despite its thinness, it integrates:

65,536 electrodes to detect neural electrical activity
1,024 simultaneous recording channels capturing brain signals
16,384 stimulation channels for sending signals back to the brain
Radio transceiver for wireless data transmission
Wireless power circuit eliminating need for internal batteries
Digital control logic managing the entire system
Data converters translating analog brain signals to digital data

Total volume: approximately 3 mm³—less than 1/1000th the size of conventional BCIs. The flexible chip conforms to the brain's curved surface, maintaining stable contact over time.

Placement: the subdural space

Neurosurgeon Dr. Brett Youngerman describes the surgical procedure: "The implants can be inserted through a minimally invasive incision in the skull and slid directly onto the surface of the brain in the subdural space. The paper-thin form factor and lack of brain-penetrating electrodes or wires tethering the implant to the skull minimize tissue reactivity and signal degradation over time."

The subdural space—between the skull's inner surface and the brain's protective membrane (dura mater)—provides an ideal location. The chip rests gently on the brain without penetrating tissue, reducing inflammation while maintaining excellent signal quality.

Data flow: from neurons to AI

Step 1 – Signal capture:

Neurons communicate using electrical impulses. When you think, move, or perceive, specific patterns of neurons fire. BISC's 65,536 electrodes detect these voltage changes across the cortical surface with high spatial resolution.

Step 2 – Wireless transmission:

The chip's integrated radio transmits data to a small wearable "relay station" using a custom ultrawideband link achieving 100 Mbps throughput. This relay station, powered by a battery, communicates with the implant and provides wireless power.

Step 3 – WiFi relay:

The relay station is itself an 802.11 WiFi device, creating a wireless network connection from any computer directly to the brain. This architecture allows real-time streaming of neural data to cloud-based AI systems.

Step 4 – AI decoding:

Advanced machine learning models analyze the high-resolution neural patterns. Trained on large datasets, these models can decode:

Motor intentions: Which movements you're planning to make
Perceptual experiences: What you're seeing, hearing, or feeling
Cognitive states: Attention, memory retrieval, decision-making
Seizure patterns: Abnormal neural activity indicating epileptic events

Manufacturing: standard semiconductor process

BISC was fabricated using TSMC's 0.13-μm Bipolar-CMOS-DMOS (BCD) technology—a versatile process integrating digital logic, high-voltage analog circuits, and power devices onto one chip. This standard manufacturing approach means BISC can be produced at scale using existing semiconductor fabs, dramatically reducing cost compared to custom medical devices.

Medical Applications: Restoring Lost Function

Epilepsy: precision seizure management

Drug-resistant epilepsy affects approximately 30% of epilepsy patients. For these individuals, medications fail to prevent seizures, and surgery to remove seizure-causing brain tissue carries significant risks.

BISC offers a middle path. Its high-resolution recording can detect the precise cortical regions where seizures originate. The stimulation channels can then deliver targeted electrical pulses to suppress abnormal neural activity before seizures fully develop.

Dr. Brett Youngerman, along with neurologist Dr. Catherine Schevon, recently secured NIH funding to implement BISC for drug-resistant epilepsy management. Early preclinical and short-term human recordings demonstrate stable signal quality—a critical requirement for long-term implantation.

Paralysis: thought-controlled prosthetics

For individuals with spinal cord injuries, ALS, or stroke, the connection between brain and body is severed. Neural commands for movement exist, but can't reach muscles.

BISC's high bandwidth enables sophisticated decoding of motor intentions. When you think about moving your arm, specific patterns fire in the motor cortex. AI models trained on these patterns can translate thoughts into control signals for:

Robotic arms and hands
Powered exoskeletons
Wheelchairs and mobility devices
Computer cursors and keyboards

The 100 Mbps bandwidth means richer, more nuanced control than previous BCIs. Instead of simple "left/right/click" commands, users could potentially control individual fingers, adjust grip force, and execute complex multi-step movements.

Speech restoration: from thoughts to words

Conditions like ALS, locked-in syndrome, and severe strokes can eliminate the ability to speak while leaving cognitive function intact. The person knows what they want to say but can't produce the motor commands for speech.

Recent BCI research has achieved over 90% accuracy in decoding attempted speech from brain signals using advanced AI. BISC's high channel count and bandwidth could accelerate this field, enabling:

Real-time speech synthesis from neural activity
Natural conversation speeds (not just slow letter-by-letter spelling)
Prosody and emotion in synthetic speech
Multi-language support

Blindness: visual neuroprosthetics

Dr. Andreas Tolias, professor at Stanford's Byers Eye Institute and co-founder of the Enigma Project, led extensive preclinical testing of BISC in the visual cortex. The goal: creating artificial vision by directly stimulating visual brain areas.

For individuals with retinal damage or optic nerve injury but intact visual cortex, BISC's 16,384 stimulation channels could create patterns of neural activity corresponding to visual scenes. A camera captures the environment; AI converts images into stimulation patterns; BISC delivers those patterns to create perception.

While full vision restoration remains years away, early applications include navigation assistance, object recognition, and reading—life-changing capabilities for the blind.

BISC vs Other Brain-Computer Interfaces

The BCI landscape in 2026 includes multiple approaches, each with different trade-offs in resolution, invasiveness, bandwidth, and target applications.

System	Electrode Type	Channels	Wireless Bandwidth	Implant Location
BISC (Columbia/Stanford)	Surface electrodes (µECoG)	1,024 recording / 16,384 stimulation	100 Mbps	Subdural (between skull and brain)
Neuralink	Penetrating threads	1,024+ recording	Unknown (proprietary)	Penetrates cortex; skull-mounted module
Paradromics Connexus	Microwire arrays (penetrating)	1,600 recording	Wired to chest module	Penetrates cortex; chest transceiver
Precision Neuroscience	Thin-film surface arrays	1,024 recording	External module	Surface of cortex
Synchron Stentrode	Vascular stent (non-surgical)	16 recording	~1 Mbps	Blood vessel near motor cortex

Key differentiators

Miniaturization:

BISC's single-chip architecture eliminates the skull-mounted titanium canisters required by Neuralink or the chest implants used by Paradromics. At approximately 3 mm³, it's the smallest fully functional BCI to date.

Bandwidth:

The 100 Mbps wireless link is at least 100× faster than competing wireless BCIs. This enables real-time streaming of high-resolution neural data to cloud-based AI systems—critical for complex decoding tasks.

Invasiveness:

Surface (µECoG) electrodes avoid penetrating brain tissue, reducing inflammation and long-term signal degradation. Synchron's Stentrode is even less invasive (inserted via blood vessels), but offers far fewer channels and lower resolution.

Scalability:

BISC's thin form factor allows multiple chips to be implanted across different brain regions simultaneously—something impractical with bulky skull-mounted systems. A patient could have one chip over motor cortex, another over speech areas, and a third over visual cortex, all wirelessly networked.

The Privacy Question: Who Owns Your Thoughts?

Unprecedented access to the mind

Every previous communication technology—writing, telephone, internet—involved conscious, deliberate transmission. You choose what to write, say, or post. BCIs fundamentally change this dynamic. They capture neural activity before conscious filtering, potentially revealing:

Intentions before actions: What you're about to do before you do it
Attention and focus: What you're paying attention to, moment by moment
Emotional states: Fear, anger, arousal, stress—potentially detectable in neural patterns
Cognitive states: Memory retrieval, decision-making processes, internal speech

Current AI can't fully decode all these states, but the technology is advancing rapidly. What happens when it can?

Who owns neural data?

No clear legal framework exists. Questions include:

Property rights: Do you own the data streamed from your neurons? Can you sell it? License it?
Third-party access: Can employers, insurers, or governments subpoena neural recordings?
Commercial use: If a BCI company trains AI on your neural data, do you deserve compensation?
Data security: What happens if neural data is hacked? Can adversaries extract passwords, PINs, secrets from brain recordings?

The Columbia team is acutely aware of these issues. Andreas Tolias states: "BISC turns the cortical surface into an effective portal, delivering high-bandwidth, minimally invasive read-write communication with AI and external devices." Read-write means bidirectional—not just reading thoughts, but potentially writing them too through stimulation.

The stimulation paradox

BISC includes 16,384 stimulation channels—far more than recording channels. While therapeutic applications are clear (suppressing seizures, restoring vision), stimulation also raises concerns:

Memory manipulation: Could stimulation alter or erase memories?
Mood control: Could external parties induce happiness, fear, or compliance?
Thought insertion: Could stimulation create perceptions or intentions you didn't originate?

Current technology can't do any of this with precision. But BISC's architecture makes these scenarios technically plausible in the future.

Regulatory vacuum

Medical BCIs fall under FDA regulation, which focuses on safety and efficacy—not privacy or autonomy. HIPAA protects health data, but doesn't address real-time neural streaming. The EU's GDPR considers neural data "biometric," offering some protection, but enforcement in this context is untested.

Experts propose new frameworks:

Cognitive liberty: A fundamental right to mental self-determination
Neural data ownership: Explicit property rights over brain-derived data
Mandatory encryption: End-to-end encryption of all neural transmissions
Opt-in AI training: Explicit consent required before neural data trains AI models
Right to disconnect: Users must be able to disable recording at will

Frequently Asked Questions

How is BISC different from Neuralink?

Three key differences: (1) Form factor: BISC is a single subdural chip (3 mm³) with no skull-mounted module; Neuralink requires a skull implant housing electronics. (2) Electrode type: BISC uses surface electrodes (µECoG) that don't penetrate brain tissue; Neuralink uses penetrating threads that insert into cortex. (3) Bandwidth: BISC achieves 100 Mbps wireless transmission—significantly higher than publicly disclosed Neuralink specs. Both target similar medical applications but use fundamentally different engineering approaches.

Is BISC safe for long-term use?

Preclinical studies show promise, but human long-term data is limited. BISC's subdural placement (no tissue penetration) minimizes inflammation compared to penetrating electrodes. The paper-thin flexible design conforms to the brain without compression. Short-term human recordings during neurosurgery confirmed signal stability and safety. However, long-term implantation studies (3–5 years) are just beginning. Key safety questions include: tissue reactivity over years, wireless power safety, signal stability over time, and infection risk. FDA approval will require extensive long-term safety data.

Can BISC actually "read thoughts"?

It can decode neural patterns, not full thoughts. BISC records electrical activity from 65,536 electrodes. AI trained on this data can decode motor intentions (which movements you plan), perceptual experiences (what you're seeing/hearing), and cognitive states (attention, memory retrieval). It cannot read complete sentences you're thinking, complex ideas, or abstract concepts—at least not yet. The AI is limited by its training data and decoding algorithms. As AI advances, decoding capability will improve, but we're far from "mind reading" in the sci-fi sense.

Who has access to my neural data?

Currently unclear—regulations don't exist yet. Medical BCIs fall under HIPAA (patient data privacy), but real-time neural streaming introduces new risks. Key concerns: (1) Can BCI companies use your neural data to train commercial AI? (2) Can employers or insurers request neural recordings? (3) What happens if your BCI is hacked? (4) Can law enforcement subpoena brain data? The Columbia team has not publicly detailed BISC's data governance model. Experts recommend: mandatory encryption, user-controlled recording, explicit consent for AI training, and new "cognitive liberty" laws protecting mental privacy.

When will BISC be available for medical use?

Likely 5–10 years for FDA approval. Timeline: (1) 2026: NIH-funded epilepsy trials beginning; (2) 2027–2029: Long-term safety and efficacy studies in human patients; (3) 2030–2033: FDA review process; (4) 2033+: Potential commercial availability. The Columbia/Stanford team founded Kampto Neurotech to commercialize BISC for research applications first, then medical use. Factors affecting timeline: trial results, manufacturing scale-up, regulatory pathway, and reimbursement negotiations with insurers.

Could BISC be used for non-medical "augmentation"?

Technically possible, but legally and ethically complex. BISC's architecture could enable: faster human-computer interaction, direct brain-to-brain communication via AI intermediaries, memory enhancement through external storage, or skill transfer through brain stimulation. However: (1) Medical BCIs are FDA-regulated for specific conditions—not general enhancement, (2) Surgical risks make elective implantation ethically questionable, (3) Long-term effects of continuous neural recording are unknown. The research team emphasizes medical applications. Non-medical augmentation would require new regulatory frameworks and extensive ethical debate.

What prevents someone from hacking a brain implant?

Critical question with no complete answer yet. Security concerns: (1) Wireless transmission could be intercepted, (2) Relay station could be compromised, (3) AI decoding systems in the cloud are hacking targets. Potential protections: end-to-end encryption of neural data, authentication before stimulation commands, local (on-chip) processing to minimize cloud dependence, and hardware-level security features. BISC's architecture includes digital control logic that could implement security, but Columbia hasn't detailed cybersecurity measures. As BCIs become common, cybersecurity will be as critical as biocompatibility.

Thinknology

Brain chips that read your thoughts