Apple has started rolling out its Audio Intelligence suite for the Apple Watch in developer beta, as part of iOS 27.3, after unveiling it at its September hardware event. The suite includes four opt-in AI features: recognition of important sounds such as sirens and doorbells, song recognition, Live Rewind and Siri Recaps. They are designed for the Apple Watch Series 12 and Ultra 4.
Live Rewind shows, after a double-press of the crown, text of what was said in the last 15 seconds, with a chime and an animation that signal to people nearby that something was captured; the snippet is deleted after 30 seconds unless saved. Siri Recaps summarizes conversations into a sentence and a few bullet points without identifying speakers, keeping raw audio or producing word-for-word transcripts. Unsaved recaps delete themselves within seven days, and the feature can be limited to a schedule.
According to Apple, raw audio is processed in a hardware-protected area, the iPhone turns it into condensed text, and the recap itself is generated in Private Cloud Compute and deleted there immediately. The conversation-listening features are the most sensitive on privacy, given the backlash against Meta's AI glasses and other recording devices.