UC Davis neuroprosthesis restores conversation for Casey Harrell

AI system

System: UC Davis intracortical speech neuroprosthesis (BrainGate2 trial)
Model: A decoder identifies phonemes. Three language models convert them into words and sentences.
Release: Exact release not established.
Application developer: UC Davis Neuroprosthetics Lab (Brandman, Stavisky) with the BrainGate consortium
Role: The models translated attempted speech into text from cortical recordings.
Configuration: Four arrays recorded signals from 256 electrodes in the left precentral gyrus. The language models used a vocabulary of 125,000 words.
Inputs and outputs: The system converted cortical activity into phonemes, then words and sentences.
Autonomy: Not assessed
Workflow: Preserved speech-motor cortical activity during attempted speech despite severe dysarthria. 256-electrode intracortical recording in the ventral precentral gyrus. Phoneme decoder plus large-vocabulary language models calibrated on the participant's own data within minutes. Regulated trial setting (BrainGate2, NCT00912041) with ongoing research support.
Human oversight: Not assessed
Relation to AI: A neural-network decoder (phoneme prediction every 80 ms) combined with language models is the mechanism converting attempted speech into text. The benefit is attributable to the decoding system, with surgery and arrays as enabling hardware.
AI contribution: A neural-network decoder (phoneme prediction every 80 ms) combined with language models is the mechanism converting attempted speech into text. The benefit is attributable to the decoding system, with surgery and arrays as enabling hardware.

Benefits

Outcome identifiers: Not assessed
Benefit: 97.5% sustained accuracy over 248 hours of self-paced conversation
Category: Improved Accessibility
Status: Not assessed
Beneficiary groups: People.
Beneficiaries: Not assessed
Reach: Not assessed
Distribution and access: Not assessed
Duration: Not assessed
Expected?: Yes
Costs and harms: Not assessed

Measurement

Studies: Not assessed
Study design: Not assessed
Measurement: Not assessed
Setting and sample: Not assessed
Measures: 97.5% sustained accuracy over 248 hours of self-paced conversation
Comparison: Not assessed
Results: 97.5% sustained accuracy over 248 hours of self-paced conversation
Additional value from AI: Not assessed
Limitations: Not assessed
Denominators: Not assessed
Materials for replication: Not assessed

Money

Kind of value: Not assessed
Amount: Not assessed
Currency and prices: Not assessed
Period and unit: Not assessed
Realization: Not assessed
Recipients and payers: Not assessed
Gross or net: Not assessed
Included costs: Not assessed
Comparison: Not assessed
Method: Not assessed
Return and payback: Not assessed

People and places

Participants: Developer: UC Davis Neuroprosthetics Lab (Brandman, Stavisky) with the BrainGate consortium. Operator: UC Davis Health.
Jurisdictions: Benefit: Davis, United States. Developer: United States. Operator: United States.
Ownership and control: Not assessed
Dependencies: Not assessed
Cross-border connections: Not assessed
Geopolitical implications: Not assessed

Follow-up

Conditions for replication: Array signal quality remains stable over months to years. Participant retains the ability to attempt speech and the cognition to operate the system. Decoder recalibration available as neural signals drift.
Replication attempts: 2 follow-up entries. Benefit recurred: 2.
Later findings: 15 June 2026. A Nature Medicine report covering nearly two years of unsupervised home use found the same participant communicated 1,960,163 words at 56 words per minute and kept full-time employment.
Availability: Not assessed
Institutional decision: Not assessed
Grounds: Not assessed
Action and result: Not assessed
Success criterion: Not assessed

Evidence

Source numbers: AIBD-0002-source-0001–0004
Citations: Numbered references below.
Language: Not assessed
Source relationships: Developer, Independent
Original evidence: Not assessed
Access and versions: Not assessed
Supported findings: Not assessed
Evidence links: Not assessed
Evidence strength: Not assessed
Missing evidence: Not assessed

Grades

Assessed benefits: People.
Depth: People: 3.
Reasons: People: Life-changing benefit to one identifiable person: conversational communication with family and colleagues was restored. The 2026 follow-up documents over 3,800 hours of unsupervised home use, nearly two million words and continued full-time employment.
Observed or intended: Not assessed
Rubric: Not assessed
Assessor and date: Not assessed
Alternative assessments: Not assessed
Fill: Independently reported

Entry details

Kind

Not assessed

Related entries

Both decode speech for people with impaired speech; Whisper fabricated words and violent phrases when transcribing impaired speech.

Field

Healthcare / assistive communication. ALS with severe dysarthria.

Purpose

Built to restore communication by decoding attempted speech from intracortical activity into text. Conversational communication was the intended benefit and is what occurred.

Dates

Date basis unassessed: 2023 (precision unassessed). Updated: 21 October 2026.

Stage

Not assessed

Version

SHA-256: 6a9f2d960f0e…

References

Other identifiers: oecd_aim: 2026-06-15-4dfc. AI Incident Database: 732.
Category mappings: Not assessed
Brief: Not assessed
Revisions: 7 revisions. Latest: 21 October 2026.

Sources

Peer-reviewed study Developer

Card NS, Wairagkar M, Iacobacci C, et al. (2024). An Accurate and Rapidly Calibrating Speech Neuroprosthesis. N Engl J Med 391(7):609-618 (issue of 15 August 2024).
Peer-reviewed study Developer

Card NS, Singer-Clark T, Peracha H, et al. (2026). Long-term independent use of an intracortical brain-computer interface for speech and cursor control. Nature Medicine.
News report Independent author

Sanders L (2026). Brain implants are helping people who've lost their voice. Science News.
Press release Developer

UC Davis Health (2024). New brain-computer interface allows man with ALS to 'speak' again. Institutional release republished by ScienceDaily.