Headlines often claim that AI can now “read minds”. The reality is less magical, but still remarkable: brain-computer interfaces (BCIs) combined with AI can turn certain brain signals into text, speech or cursor movements, helping people who have lost the ability to move or speak.
They cannot, however, read your private thoughts at a distance. Here is what this technology can really do today.
Key Takeaways
- A BCI records brain activity and translates it into commands for a computer or device
- The most powerful BCIs use electrodes implanted in or on the brain
- Implanted speech BCIs have helped paralyzed people communicate at dozens of words per minute
- Non-invasive decoders exist but need the person’s cooperation and hours of training
- Laws protecting “neural data” are starting to appear
What Is a Brain-Computer Interface?
A brain-computer interface is a system that records brain activity and translates it into commands. It works in four basic steps:
- Sensors record electrical (or blood-flow) activity from the brain
- The signal is filtered and amplified
- AI and machine learning models learn which patterns match which intentions
- The system turns those patterns into actions, such as moving a cursor or producing a word
Invasive vs. Non-Invasive BCIs
| Type | How It Works | Pros and Cons |
|---|---|---|
| Implanted electrodes | Placed in or on the surface of the brain by surgery | Most precise signals, but require brain surgery |
| Endovascular | A device placed inside a blood vessel near the brain | No open-brain surgery, but less detailed signals |
| EEG headsets | Electrodes on the scalp | Safe and cheap, but noisy, limited signals |
| fMRI | Measures blood flow in a large scanner | Detailed maps, but slow and not portable |
Key Breakthroughs
Moving Cursors and Robotic Arms
The BrainGate research consortium has worked for about two decades with people with paralysis, showing that implanted electrodes can let them move computer cursors and robotic arms with their thoughts.
Turning Brain Signals Into Speech
In 2023, the journal Nature published two major studies. A Stanford team helped a woman with ALS communicate at about 62 words per minute through an implanted BCI, and a team at the University of California, San Francisco, helped a woman paralyzed by a stroke speak through a digital avatar at around 78 words per minute. In 2024, a UC Davis team reported that a man with ALS could communicate with about 97% accuracy using an implanted speech BCI.
Decoding Meaning Without Surgery
In 2023, researchers at the University of Texas at Austin used fMRI and an AI language model to reconstruct the general meaning of stories people heard or imagined. However, the decoder needed many hours of training for each person, only captured the gist rather than exact words, and failed when the person did not cooperate.
Companies in the Field
- Neuralink: implanted its first device in a human patient in early 2024, allowing him to control a computer cursor with his thoughts
- Synchron: develops the Stentrode, a device inserted through a blood vessel rather than through open-brain surgery
- Blackrock Neurotech: makes the Utah array electrodes used in much academic BCI research
- Precision Neuroscience: develops a thin electrode film placed on the surface of the brain
The Voynich Manuscript: Facts, Theories and Decoding Attempts
Real-World Applications
- Communication: helping people with ALS, stroke or spinal cord injuries to type or speak
- Mobility: controlling wheelchairs, robotic arms or prosthetic limbs
- Medicine: research on epilepsy, depression and other brain disorders
- Consumer devices: EEG headbands for meditation, sleep or focus, with much more limited capabilities
Limits and Ethical Questions
Current BCIs are still mostly experimental. Implanted devices require surgery, each system must be trained for each person, and they work best for specific tasks, such as attempted speech or movement, not for reading random thoughts.
Still, neural data is extremely personal, and experts warn about privacy as consumer devices spread. In 2021, Chile became the first country to amend its constitution to protect “neurorights”, and in 2024 Colorado passed a law to protect neural data under its privacy law.
FAQ
Can AI really read your mind?
Not in the science-fiction sense. AI can decode specific brain signals, like attempted speech or movement, mostly with implanted devices and the person’s cooperation. It cannot read private thoughts at a distance.
What exactly is a brain-computer interface (BCI)?
It is a system that records brain activity and translates it into commands for a computer or device, such as moving a cursor or producing text.
Can BCIs help people with disabilities?
Yes. Research studies have helped people with paralysis move cursors and robotic arms and communicate through text and synthetic speech.
Are there privacy concerns with brain-computer interfaces?
Yes. Neural data is very sensitive, which is why Chile and some U.S. states have begun passing laws to protect it.
Can I buy a mind-reading device today?
Consumer EEG headbands exist for meditation, sleep or focus tracking, but they can only detect general brain states. Advanced implanted BCIs are still limited to clinical trials.