Meta Version 2 BCI Translates Brain Activity Into Keypresses via MEG

By Tom Whitfield

Meta has released version 2 of a non-invasive brain-computer interface that converts brain activity into keypresses through a magnetoencephalography (MEG) scanner, Tom's Hardware reported. The system represents the company's second public iteration of a thought-to-keypress device that operates entirely outside the skull.

What modality does the version 2 system use?

The decoder relies on MEG, a recording technique that captures the weak magnetic fields produced by postsynaptic neuronal currents. Unlike scalp EEG, MEG does not require conductive gel, skin abrasion, or electrode arrays mounted on the scalp. Unlike intracranial ECoG or implanted electrode arrays, it requires no craniotomy, no dural breach, and no chronic intracranial hardware.

The Meta scanner detects shifts in brain activity and routes those signals through a decoding algorithm that outputs a single keypress per interpreted intention on a standard keyboard. The Tom's Hardware summary does not disclose the channel count, sampling rate, classifier type, vocabulary, or decoding latency of the version 2 release.

How does the thought-to-keypress pathway work?

The user's intent to type registers as a pattern of magnetic flux changes around the head, captured by superconducting sensors housed in a helmet-shaped array. The algorithm translates those spatiotemporal patterns into discrete keypress events, allowing text entry without finger movement. The application aligns with proof-of-concept BCI typing demonstrations that have appeared in the research literature for imagined motor actions and language-related cortical activity.

The version 2 label confirms at least one prior public release from Meta's BCI program. The Tom's Hardware coverage does not detail the engineering differences between the two iterations, nor does it report decoding accuracy, character error rate, vocabulary size, or signal-to-noise benchmarks that would allow direct comparison between v1 and v2.

What is the regulatory status?

Meta's release sits within the investigational category. The Tom's Hardware report contains no reference to FDA clearance, a clinical trial registration number, or a regulatory submission pathway. The MEG-based architecture does not match any FDA-cleared BCI device on the US market, and clinical adoption of the version 2 platform would require a separate regulatory review process before any patient-facing deployment.

The release also lacks reference to an Investigational Device Exemption, an IRB approval, or any peer-reviewed validation cohort.

What does this mean for clinical neurophysiology?

For laboratories accustomed to scalp EEG and intracranial monitoring, the version 2 release reinforces MEG as a viable BCI signal source at a time when implanted arrays dominate clinical headlines. MEG provides higher spatial resolution than scalp EEG for deeper cortical sources and superior temporal resolution to fMRI, though it demands superconducting quantum interference device (SQUID) sensors, magnetic shielding rooms, and fixed head positioning that constrain experimental throughput.

The system's non-invasive profile removes the surgical, anticoagulation, hardware-infection, and post-operative imaging surveillance risks associated with implanted arrays. The trade-off is hardware footprint: MEG installations require dedicated shielded enclosures with helium-cooled detector helmets, restricting deployment to specialist centers.

What comes next?

The release positions Meta to publish decoding performance metrics for the version 2 system against its predecessor and against competing non-invasive approaches. Watch for any preprint, peer-reviewed paper, or conference demonstration that quantifies keypress error rate under controlled conditions.

For now, the version 2 BCI adds another data point to a non-invasive BCI category that continues to develop in parallel with the implanted half of the technology space, and clinical neurophysiology practitioners will want performance numbers before any reading-room or monitoring-unit application can be considered.