FDA Clears Natural Cycles’ AI-Updated Birth Control Algorithm

Natural Cycles said on July 27, 2026 that US regulators have cleared a rebuilt version of the algorithm behind its birth control app, the sixth clearance the company has collected since 2018, when it became the first app the Food and Drug Administration authorized as a method of contraception. The point of the update is not better protection against pregnancy. It is fewer days on which the app tells users they need any.

The newly cleared software pairs statistical modeling of the menstrual cycle with machine learning the company says was trained on tens of millions of real-world fertility data points. Natural Cycles reads body temperature, taken with a basal thermometer or a wearable, and returns a daily verdict: a red day, when pregnancy is possible, or a green day, when the app says protection is not needed. The retrained model is meant to return more green days for the same user, from the same underlying biology.

Natural Cycles says effectiveness holds where it has been since the original clearance, at 98% with perfect use and 93% with typical use, and that the new model handles irregular cycles and noisier temperature readings better than its predecessor. What the company has not said is by how much. The announcement carries no figure for the additional green days, no clearance number, and no published data behind the claim that effectiveness survived the rewrite.

What the published evidence covers

The company’s research team does publish. A paper that appeared in June 2026 in The European Journal of Contraception & Reproductive Health Care, written by a group that includes both founders, the company’s senior medical advisor and outside academic co-authors, compared pregnancy and continuation rates for the app when temperature arrives from a wearable against temperature entered from an oral thermometer.

That is a real result, and it answers a different question. Whether wearable input degrades performance is not the same question as whether a retrained model can widen the non-fertile window without letting more pregnancies through. On the second question, the public record today consists of the agency’s decision and the company’s characterization of it.

The distinction matters more here than in most software. A green day is an instruction to a user, and the cost of the model being wrong is a pregnancy. Widening the window is the kind of change where the effect size is the entire story, and it is the number that has not been released.

Six clearances, and the boundary this one draws

The clearance history is the more durable asset. Natural Cycles won a De Novo authorization in 2018, the first the agency granted a contraceptive app, then cleared Oura Ring integration in 2021 and Apple Watch in 2023. A 2024 clearance came with what regulators call a predetermined change control plan: a pre-agreed envelope of future modifications a manufacturer can ship without returning to the agency. That plan is what let the company launch its own NC° Band wristband and add Garmin watches in March 2026. A 2025 clearance opened both over-the-counter and prescription supply.

The new decision marks where that envelope ends. Hardware integrations flow through the change-control plan. Rewriting the core algorithm did not, and required a fresh submission instead. For a company whose product is software that learns from an accumulating dataset, that boundary sets the pace at which the learning can reach users.

Most cycle-tracking apps never took this road, staying on the wellness side of the line where no clearance is required and no effectiveness claim can be made. Clearance is the slower trade, and it is the same one other clinical-AI developers have made, from diagnostic pathology to digital autism screening.

The proxy underneath all six clearances

Every one of those authorizations rests on the same premise: that basal body temperature is a usable stand-in for hormones. Temperature rises after ovulation because progesterone rises. It is a downstream readout, not a measurement.

That premise now has an institutional challenger. Germany’s federal breakthrough-innovation agency has opened a €40 million challenge for continuous hormone monitoring, and its rules are pointed. Teams must capture at least four hormones from a panel including estrogen, progesterone, LH and FSH, directly and on-site, over at least seven consecutive days in a fluid such as interstitial fluid, sweat or saliva. Surrogate signals are disqualified outright, with heart rate and temperature named. The program was launched by Ida Tin, the Clue co-founder who coined the term femtech, and applications closed in June 2026. Capital is moving the same direction in health hardware more broadly, including XPANCEO’s $250 million raise for an AI-powered smart contact lens.

Elina Berglund Scherwitzl, the CERN physicist who co-founded Natural Cycles and runs it, is unbothered by the exclusion. Temperature, she told Forbes, is “a remarkably powerful biomarker because it reflects the body’s physiological response to progesterone after ovulation.” She said the company has never been tied to a single biomarker and is studying whether estrogen could flag the start of the fertile window, and that she would fold continuous hormone sensing in if it becomes practical and affordable enough for daily use.

The update reaches an installed base the company puts at more than six million registered users. The measure that will show what the new algorithm delivers is straightforward: how many additional green days it grants, and at what pregnancy rate. Natural Cycles has published numbers like that for earlier versions of the algorithm.