Apple’s most ambitious AI health project is not ready for prime time. At WWDC 2026, the company confirmed it is delaying “Project Mulberry,” its internally developed AI health-coaching initiative, even as it moves forward with a narrower but still notable AI feature: sleep-sound detection that can identify snoring, coughing, and sleep-talking through a new “Sound Report” in the Health app. The features are expected to ship alongside the Apple Watch Series 12, which adds 5-second heart-rate sampling, a meaningful jump in monitoring frequency from prior generations.
What Project Mulberry Was Supposed to Be
Project Mulberry has been described as Apple’s answer to the AI health-coaching push seen across the industry — a system meant to synthesize a user’s Apple Watch and iPhone health data into personalized, conversational guidance, similar in spirit to what Google is attempting with its new Gemini Health Coach and what Samsung and Whoop are building into their own platforms. Apple’s decision to delay it at its own flagship developer conference, rather than quietly slip the timeline, signals that the company judged the feature not ready, likely due to the accuracy and liability bar required for a coaching product that touches millions of iPhone users’ health decisions.
The Feature That Did Ship: Listening While You Sleep
Instead, Apple’s near-term health AI push centers on passive audio-based detection during sleep. The new Sound Report uses the iPhone or Apple Watch’s microphone to identify and log snoring, coughing, and sleep-talking overnight, compiling the results into a health record a user can review or eventually share with a physician. Unlike Project Mulberry’s ambition to actively coach, this feature simply detects and reports — a far more conservative but more immediately deployable use of AI in the health stack.
Series 12’s Hardware Leap: 5-Second Heart-Rate Sampling
The Apple Watch Series 12, expected the following month, pairs with these software features by increasing heart-rate sampling frequency to every 5 seconds, a substantial increase in monitoring granularity compared to earlier Apple Watch generations that sampled less frequently. That density of data is precisely what makes features like sleep-sound correlation and future coaching tools more viable, since anomalies can be tied to much more precise physiological timestamps.
Caution Versus Competitive Pressure
Apple’s delay stands out against a backdrop of rivals moving fast: Google is launching a paid AI health coach the same season, and Samsung is rolling out AI-driven biomarkers like an “AGEs Index” for aging. Some analysts view Apple’s caution favorably, arguing that a coaching product touching millions of users’ health decisions carries far more downside risk if it gives bad advice than a simple detection feature does, and that Apple is right to prioritize reliability over speed. Others see the delay as a symptom of Apple falling behind in the generative AI race more broadly — the company has faced criticism throughout 2025 and 2026 for slower AI feature rollouts compared to Google and OpenAI-linked competitors, and shelving its flagship health AI product at its own showcase event reinforces that narrative.
Why the Distinction Matters
The gap between “detect and report” and “actively coach” is not trivial. Passive detection tools like Sound Report carry lower regulatory and liability risk because they simply surface information a user or their doctor can act on. Active coaching tools, like Project Mulberry or Google’s Gemini Health Coach, take on a more direct role in influencing health behavior and decisions — a threshold that invites more scrutiny from regulators, physicians, and users themselves if the advice proves wrong. Apple’s choice to ship the safer feature while delaying the riskier one suggests a deliberate, if slower, strategy.
What’s Next
Apple has not given a firm new timeline for Project Mulberry, leaving open the question of whether it resurfaces later in 2026 or slips further. In the meantime, the Sound Report feature and Series 12’s denser heart-rate data give Apple a foundation to build on: better sleep and cardiac data collected now could make a future coaching product, whenever it ships, considerably more capable. The bigger industry question is whether Apple’s more cautious approach becomes a competitive liability as Google, Samsung, and Whoop all push AI coaching products to market in the same year — or whether Apple’s reputation for reliability ends up being the safer long-term bet in a category still working out its regulatory footing.
