Brain Implants Can Win Approval Faster Than They Can Prove Value

The central tension in brain-computer interfaces is that approvals, funding and home use are all moving faster than the evidence, the payment systems and the privacy protections that everyday use depends on. Our signal scan found the devices reaching the market first. This part looks at where the evidence pulls in opposite directions.

Key points

• China approved an invasive BCI for use outside a trial and gave it an insurance code within days, while the US still has no Medicare coverage path for implants.
• One user has run a speech neuroprosthesis at home for two years at 99% word accuracy, but scaling now rests on a single participant and company press releases.
• Neural data laws are spreading across states and through UNESCO, but the enacted versions mostly lack enforcement.
• Consumer EEG earbuds can collect brain signals outside health privacy law, and short recordings can identify the person wearing them.
• What settles these questions is US payer decisions, peer-reviewed results from many participants, and the first enforcement action over neural data.

The tensions

Is regulatory approval the finish line for implants?

Approval looks close to routine. China's regulator cleared Neuracle's NEO in March 2026 and assigned it a health-insurance code days later, the first step toward state reimbursement¹. China now has five approved BCI products and 134 registered trials². In the US, Synchron raised $200M for a pivotal trial aimed at the first premarket approval for a permanent implant³, and Neuralink says it will move to high-volume production in 2026⁴.

The case against is about payment and proof. A former CMS chief medical officer warns that Medicare has no coverage framework for these devices. Trials that measure cursor control may not show the health outcomes CMS requires, so approved implants could end up reaching only self-paying or trial patients⁵. NEO is an eight-electrode device with modest functional gains, so its approval may set a low evidence bar¹. Neuralink's production targets come from social media posts, and no premarket approval has been granted⁴.

What would settle it: whether CMS sets a coverage pathway before Synchron files, and whether the pivotal trials measure function rather than device control.

Is speech decoding proven enough to scale?

The strongest case comes from home use. One man with ALS used an implanted speech BCI for over 3,800 hours across nearly two years, at 56 words per minute and 99% word accuracy without researcher setup, and kept a full-time job⁶. Independent home use for both speech and computer control is now in the peer-reviewed literature⁷. Paradromics reports that a woman with motor neuron disease was holding open-ended conversations within three months of her implant⁸.

The weaker side is the size of the evidence. The flagship result comes from one participant⁶. Paradromics announced its progress in company releases before peer review, and its durability data will take six years to arrive⁸. When a language model fills in uncertain phonemes, it is not always clear whose words come out⁶.

What would settle it: peer-reviewed results from several participants across at least two device makers, with error rates broken out for decoder guesses and decoder corrections.

Do neural privacy laws protect anyone?

Rules are spreading. Vermont became the fifth US state with a neural-data statute, and bills are pending in at least six more⁹. UNESCO adopted the first global ethics standard for neurotechnology and gave neural data heightened protection¹⁰.

Their force is thin. Vermont dropped the consent requirement, the private right of action and the penalties before passing its law⁹. UNESCO's text is non-binding and does not say who owns neural data¹⁰. Meanwhile, the same company can sell an FDA-cleared in-ear EEG and a consumer earbud that sits outside HIPAA¹¹. A study found that a consumer headset can identify individuals from short recordings regardless of mood, which undercuts promises to de-identify the data¹². Decoders that need only 10 minutes of calibration per user¹³, and an ultrasound interface funded mainly by OpenAI and aimed eventually at the general public¹⁴, increase what that data can reveal.

What would settle it: the first enforcement action over neural data, or a federal standard that includes consent and penalties.

Why these questions matter

While these questions stay open, we would treat the approval count as a weaker indicator than coverage decisions and replicated results. That applies to investors, clinicians and policymakers alike. We think it is likely, at about 70%, that the first US implant approval arrives before a Medicare payment route does. Anyone building or buying neurotech should assume their neural data falls under the device maker's terms of service until a regulator enforces otherwise.