AI tools can translate some brain activity into words and images

In short: Researchers are getting better at using AI to turn certain brain signals into simple words or images, but it still requires cooperation and controlled setups.

What's going on

Several recent studies show steady progress in “brain decoding,” which means translating patterns of brain activity into something readable, like text or pictures (like turning a faint radio signal into clearer sound). These systems are sometimes described as “mind-reading,” but experts say that wording can make the tech sound more powerful than it is.

In a 2025 Stanford study, researchers used an implanted brain-computer interface, which is a device placed in the body that can record brain signals. With AI, the system decoded imagined sentences in real time, reaching up to 74% accuracy in a limited task where the choices were constrained. The researchers said open-ended decoding, meaning trying to interpret whatever a person is thinking, was much less reliable.

In 2026, another team used brain scans to infer what images people were looking at and then recreate those images. The goal is to help people with severe paralysis communicate. Reconstructing dreams or unrestricted thoughts was described as a future aim, not something the system can do today.

What to watch

Privacy and consent are becoming central issues. In the implant study, researchers used a “mental password” so decoding only worked when participants deliberately activated it. Even so, the work suggests brain signals can include more than the exact words someone was asked to imagine, which is why researchers and reviewers are calling for stronger safeguards before these tools move beyond the lab.

Source: NYTimes