Apple launches beta for always-listening AI tools on Apple Watch

New ambient AI features

Apple has released the first developer betas for iOS 27.3 and watchOS 27.3, introducing a suite of tools known as Audio Intelligence. These features allow the Apple Watch to listen to and process nearby speech in an ambient fashion to provide summaries and quick references. The new tools are currently being tested on the Apple Watch Series 12 and the Apple Watch Ultra 4.

The hardware requirements for these features are specific, as they rely on the S11 chip. This chip includes a Secure Exclave, which is a dedicated hardware-protected area designed to handle sensitive audio data in isolation. To use these tools, the watch must be paired with an iPhone 16 or newer that supports Apple Intelligence and Siri AI, though the iPhone 16e is excluded from this compatibility.

These features were first revealed during Apple's September "Surprise and Shine" hardware event. While they are currently in a developer beta, a public beta is expected to follow later this year. Apple is targeting a full consumer release for the features in December, though some behaviors and language support may still change during the testing phase.

Siri Recap and Live Rewind

One of the primary tools in this update is Siri Recap, which uses ambient listening to detect conversations the wearer is involved in. Once a conversation ends, Apple Intelligence generates a title, a short summary, and a list of key points within the Siri app. It is important to note that the system does not provide a full transcript or identify different speakers during this process.

Users have several ways to control how Siri Recap functions, including the ability to schedule it for specific times, such as only during work hours. The feature can also be toggled on or off via the Control Center. After a recap is generated, users are prompted to review the information and can choose to save it; otherwise, the recaps are automatically deleted after seven days.

The second major tool is Live Rewind, which maintains a short rolling buffer of audio. By double-pressing the Digital Crown, a user can see a text snippet of roughly the previous 15 seconds of speech. This is intended for quick reference, such as catching a restaurant special or hearing directions again. Even when the watch is in silent mode, the device will play an audible chime and show an on-screen animation when this action occurs.

Privacy and data handling

Apple has implemented several layers of privacy protection to handle the continuous nature of these tools. Raw audio is processed inside a dedicated hardware-protected buffer on the S11 chip, which means the audio is not saved as a file. This buffer is inaccessible to the rest of the operating system, third-party apps, or Apple itself.

For the summaries generated by Siri Recap, the condensed text is encrypted before being sent to Private Cloud Compute for processing. Once the summarization is complete, the intermediate data is deleted. If a user chooses to save a snippet or a recap, that content is synced via iCloud using end-to-end encryption, provided the user has two-factor authentication and a device passcode enabled.

The system is also designed to automatically omit sensitive information from summaries. This includes things like financial data or government identifiers. By focusing on condensed text and short buffers rather than persistent recordings, Apple aims to balance the utility of ambient AI with user privacy. The features are strictly opt-in and are currently limited to the English language.

Legal and regional limits

The rollout of these features comes at a time of significant industry debate regarding AI wearables that listen to ambient audio. Some observers have pointed out potential conflicts with two-party consent eavesdropping laws in various U.S. states. Because bystanders may not receive a clear notification that their speech is being processed, these tools face a complex legal landscape.

Apple's approach attempts to address these concerns through visible and audible cues. For example, Live Rewind provides a chime and animation to signal that the audio buffer is being accessed. Additionally, the fact that the system does not create persistent raw recordings is a key part of their design strategy to differentiate from other recording devices.

There are also geographic and age restrictions on the new tools. The features will not be available in the European Union initially because broader Siri AI support is not yet available in that region. Furthermore, the tools are not available to users under the age of 13, and some reports suggest they may be restricted to users under 18 as well.

What to expect next

As the developer beta continues, users can expect to see refinements in how the AI identifies the end of conversations. The goal is to ensure that Siri Recap provides accurate summaries without capturing irrelevant background noise. Apple will likely monitor the battery impact of these features, though they have currently described the impact as negligible.

The transition from developer beta to public beta will be the next major milestone for these tools. During this phase, Apple may adjust the language support or the specific types of data the AI is allowed to summarize. The final consumer release in December will be the first time the general public can test these ambient tools on their own devices.

For those interested in the technical side of these updates, Apple has published a technical privacy overview and a support article. These documents provide deeper insight into how the S11 chip handles the Secure Exclave and how the Private Cloud Compute interacts with the watch. This transparency is intended to help users understand how their data moves between the device and the cloud.

Technical specifications

The hardware requirements for these features are quite specific. Users must own an Apple Watch Series 12 or an Apple Watch Ultra 4 to access the new tools. These specific models are required because they house the S11 chip, which provides the necessary power and security architecture for the new AI features.

In addition to the watch hardware, the paired iPhone must be an iPhone 16 or newer. This is because the iPhone must support Apple Intelligence and the underlying Siri AI framework. Notably, the iPhone 16e is excluded from this list, meaning it will not be able to facilitate these specific watch features.

While the features are currently in a testing phase, the core architecture is set. The use of the Secure Exclave for isolated audio handling is a permanent hardware feature of these watches. This ensures that even as software updates arrive, the fundamental privacy boundary between the raw audio and the rest of the device remains intact.