There's a real, currently FDA-approved medical device that continuously reads a patient's blood sugar and automatically decides how much insulin to release into their bloodstream, adjusting every five minutes, with no human approving each individual dose. It's not a hypothetical future scenario. It's standard-of-care diabetes treatment, right now.
Confirmed
Five FDA-approved automated insulin delivery systems — MiniMed 780G, Tandem Control-IQ, Omnipod 5, iLet, and CamAPS FX — are commercially available in 2026, and the American Diabetes Association's 2026 Standards of Care designated automated insulin delivery as the preferred method for all type 1 diabetes patients and insulin-requiring type 2 diabetes adults, the first time this level of recommendation has been given for the type 2 population specifically. These "hybrid closed-loop" systems combine a continuous glucose monitor, an algorithm — using techniques ranging from PID control to model predictive control to AI-enhanced prediction — and an insulin pump, automatically adjusting delivery based on real-time readings. Clinical studies show users achieving over 70% time in their target glucose range with under 2% time in dangerous hypoglycemia, and average HbA1c around 7%, alongside significant reductions in dangerous blood sugar episodes compared to manual insulin management.
The field is actively moving toward fully closed-loop systems that eliminate manual mealtime input entirely — no carbohydrate counting, no bolus announcements — relying purely on algorithm-detected glucose changes. One feasibility study of a fully closed-loop system in adults with type 2 diabetes achieved 68% time in range with zero manual meal interactions. Beta Bionics' iLet system has simplified setup to require only the patient's body weight to initiate therapy. Separately, in early 2026, the first Canadian ALS patient received a Neuralink brain-computer implant — a different category of AI-mediated medical device, one that reads neural signals rather than managing a metabolic process, but part of the same broader trend of AI systems taking on real-time physiological decision-making that used to require a human clinician's judgment.
Genuinely uncertain
A specific, documented safety concern with these systems is real and worth naming plainly: because fully CGM-controlled closed-loop systems depend entirely on the accuracy of the glucose sensor feeding them data, a false CGM reading can cause the algorithm to deliver an insulin overdose, risking severe hypoglycemia. Researchers are actively developing safeguards against exactly this failure mode — one 2026 published design proposes a dual closed-loop system specifically to add a second layer of protection against a single sensor's error propagating into a dangerous automated dose. That this is an active area of engineering research, not a solved problem, is itself the honest state of the science: the risk is understood and being actively mitigated, not eliminated.
Worth examining honestly
PROMETHEUS would tell you this is exactly the kind of AI application worth being unambiguously enthusiastic about — a measurable, FDA-vetted improvement in a chronic disease that used to demand constant vigilance and carried real overnight risk from missed corrections. CIPHER-7 would add that "FDA-approved" and "risk-free" are not the same claim, and the sensor-accuracy failure mode is a legitimate reason these systems require ongoing clinical supervision, not a reason to distrust them outright — the same caution that applies to any automated system making consequential decisions with limited human oversight in the loop.
The Door This Opens
This door isn't a hypothetical about AI "controlling your body" someday. It's already happened, for a specific, well-documented, and genuinely beneficial medical use case — and the honest lesson in how it happened responsibly is worth remembering as more categories of AI-mediated physical intervention move from research to approval: rigorous clinical trials, transparent failure-mode research, and a regulatory body willing to require both.
This article discusses FDA-approved medical technology for general audience education and is not medical advice. Anyone managing diabetes should consult their healthcare provider about treatment options.