Researchers at the National University of Singapore and Nanjing University of Posts and Telecommunications have described NeuroAxis, a vagus-nerve implant that is “solely powered by endogenous glucose, operating without batteries or wireless power transfer.” The findings are from rat models of type 1 and type 2 diabetes only. They appear in a Nature paper published on 18 November 2026, “A glucose-powered artificial neuroimmune axis for metabolic control,” and in an NJUPT release the following day.
The work has not been validated in people and has not been cleared by any regulator. It is an animal study, not a treatment for diabetes in humans.
The paper says NeuroAxis “emulates the vagal reflex through closed-loop neuromodulation with metabolic interoceptive feedback.” In the authors’ words: “In a rat model of type 1 diabetes, NeuroAxis promotes the recovery of insulin-producing pancreatic β-cells and reduces inflammatory immune responses; in a rat model of type 2 diabetes, it drives macrophage remodelling and reduces systemic chronic inflammation.” “In both models, NeuroAxis restores glucose tolerance, improves insulin sensitivity and normalizes glycaemic control, with therapeutic effects persisting after treatment cessation.” Those results, including the phrase about effects after “treatment cessation,” are the paper’s account of what happened in rats. They are not evidence that the implant has been shown to work in patients.
The corresponding authors are Yuxin Liu of NUS and Lianhui Wang of NJUPT. The NJUPT release names Weijun Xiu, an NJUPT alumnus, and Qian Wang, a doctoral student at NUS, as first authors, which matches the Nature author list.
NJUPT’s release, in a TSN translation, says NeuroAxis is made of three core modules: a glucose fuel cell, a flexible logic-control circuit and a deformable nerve electrode. The fuel cell, based on nanozyme catalysis, “converts glucose in the body into electrical energy and, at the same time, uses changes in its output voltage to sense glucose concentration,” combining power and sensing in one unit. When the system detects a sustained high-glucose state, it starts electrical stimulation of the vagus nerve; when glucose returns to normal, it stops stimulating and goes back to monitoring. The release does not, in that description, publish a voltage threshold, and TSN is not adding one.
NJUPT also says Nature ran a News & Views commentary on the paper, under the title “A step closer for implants that treat diabetes.” That headline is Nature’s. It should not be read as a finding that the implant treats diabetes in people. The paper’s own code-availability statement says the custom code for the closed-loop control logic is on GitHub.
A result in diabetic rats sits in a different category from studies already done in patients. A Johns Hopkins-led study of 397 young people with type 1 and type 2 diabetes found that almost all of those offered an autonomous AI eye exam at a clinic visit completed the screening. That was a human study of a screening tool, not an implant, and it does not make NeuroAxis a clinical device. Closer in kind is TSN’s recent coverage of a Nature paper on an UNC13A splice correction that restored signalling in laboratory motor-neurone-disease models, with no patient data. Lab models can be careful and still be only lab models. NeuroAxis belongs on the same side of that line, and in this week’s Tech for Good round-up of laboratory and public-health findings.
A result in rats is not a treatment, and it is not a clearance
The paper’s strongest sentences are also the ones that have to be kept inside their fence. NeuroAxis, the authors say, is powered only by the body’s own glucose, needs no battery and no wireless power, and in rat models of type 1 and type 2 diabetes it restored glucose tolerance, improved insulin sensitivity and normalised glycaemic control, with effects that continued after what the paper calls “treatment cessation.” In the type 1 model it promoted recovery of insulin-producing beta cells; in the type 2 model it remodelled macrophages and reduced chronic inflammation. NJUPT’s account of the three modules, and of stimulation that starts on sustained high glucose and stops when levels normalise, is a description of how the team says the closed loop is meant to work in those animals.
None of that is a human trial, a named patient, or a regulator’s decision. The abstract’s language of “therapy” and “treatment,” and Nature’s News & Views headline about implants that “treat diabetes,” are the authors’ and the journal’s wording. They do not convert a rat study into a cleared treatment. Until a clinical study or a regulatory filing is published, the accurate reading is the narrow one: a glucose-powered vagus-nerve implant has been tested in diabetic rats, the authors report the metabolic and immune changes above, and the device has not been validated in people. The next facts to watch are a larger-animal study, a registered clinical plan, or a regulator’s review. None of those appear in the paper or the NJUPT release.
How we sourced this
The quoted findings are taken verbatim from the Nature abstract, as set out in Research’s brief. Details of the three modules and of when stimulation starts and stops are TSN translations from NJUPT’s Chinese release. No voltage threshold or other figure that is absent from the abstract and the release has been added.
Sources
- Weijun Xiu, Qian Wang et al., “A glucose-powered artificial neuroimmune axis for metabolic control,” Nature, 18 November 2026.
- Nanjing University of Posts and Telecommunications, “NJUPT Professor Lianhui Wang and a National University of Singapore team publish the latest research findings in the field of autonomous bioelectronic therapy in Nature,” 19 November 2026 (Chinese; TSN translation of the official title).
Get TSN’s biggest tech stories, explained.
News and insights across AI, tech, crypto and cybersecurity. Every source linked. Unsubscribe anytime.
Get the newsletterRelated stories
UNC13A Fix Restores Neuron Signalling in Lab MND Models is the closest recent TSN account of a carefully reported laboratory finding that has not been tested in patients, and is the next read.
- AI Eye Exams Closed a Youth Diabetes Screening Gap, Study Says
- Tech for Good This Week: Blood Tests, Methane Leaks and Lab Models
Follow @tsnmediahub on X for breaking AI news as it happens.
Follow on X

