Google has started enabling its “Sensitive Content Warnings” feature within the Android Messages app, delivering on a capability first detailed in October 2024. This system employs artificial intelligence running locally on user devices to detect potential nudity in images, automatically blurring them and offering controls to both recipients and senders before content is displayed or sent.
The rollout marks the culmination of development that saw backend components distributed months prior, and arrives amidst broader industry efforts to combat harmful online content and increasing AI-driven cyber threats.
The feature introduces a two-pronged approach to handling potentially explicit images within the app, which supports modern Rich Communication Services (RCS) messaging alongside traditional SMS/MMS. When an incoming picture is flagged by the on-device AI, it appears blurred in the conversation. The recipient sees options to learn more about the potential harms associated with such content or block the sender.
To proceed, they tap “Next,” then choose between “No, don’t view” or “Yes, view,” according to 9to5Google. If viewed, an option remains to “Remove preview” and re-blur the image. A parallel warning system prompts users who are *sending* an image identified as potential nudity, advising them to reconsider.
To send the image, they must tap “Next” and then explicitly confirm by swiping “Yes, send,” or cancel by selecting “No, don’t send”. Currently, this detection applies only to images and does not yet analyze videos.
On-Device Processing and Privacy Assurances
Central to the feature’s design is its reliance on local processing to maintain user privacy. The image analysis is handled by a system component known as Android System SafetyCore, a service providing “privacy-preserving on-device support for Android safety features,” according to Google’s support documentation.
This component started rolling out via Google System Updates in November 2024, alongside another app called “Android System Key Verifier” (intended for future contact identity verification). G
Google emphasizes that image classification runs “exclusively on your device and the results aren’t shared with Google.” Furthermore, the official documentation clarifies that SafetyCore “is only active when an application integrates with SafetyCore and specifically requests that content be classified. Images aren’t classified unless they’re sent through Google Messages with the Sensitive Content Warnings feature turned on.”
User Reactions and System Rollout
Despite these privacy assurances, the earlier, often unnoticed installation of the SafetyCore component raised some user concerns, with some prematurely labeling it due to the lack of initial transparency about its purpose before the feature was officially active. The component operates solely on the device, a point echoed by technical analyses.
The user-facing Sensitive Content Warnings feature within Messages is currently becoming available to those in the Messages beta program. Google’s original announcement indicated a broader rollout “to Android 9+ devices including Android Go devices with Google Messages in the coming months.”
Android Go is a lighter version of the operating system designed for entry-level hardware. The SafetyCore app description in February 2025 also noted a “gradual rollout in 2025,” aligning with the current activation timeframe. The feature requires devices to have at least 2GB of RAM. Users can find the control toggle within Messages by navigating to Settings > Protection & Safety > Manage sensitive content warnings.
As with many AI classification systems, Google acknowledges that the feature may occasionally misclassify images, potentially flagging safe images or missing explicit ones.
Managing Warnings and Availability
How the feature behaves by default depends on the user account configuration. For most adults using unrestricted phones, the warnings are off by default and must be actively enabled (opt-in). For teenagers (13-17) whose accounts are not supervised through Family Link, the feature is enabled by default but can be turned off in their Google Account settings.
For children whose accounts are supervised via Family Link (Google’s parental control service), the feature is enabled by default, and only the parent or guardian can disable it through the Family Link app.
This nudity warning system arrives as part of a wider push by Google to enhance safety within Messages using AI. The October 2024 announcement also covered enhanced scam detection capabilities focusing on specific patterns like fake job or delivery notifications, which began reaching Pixel phones in March 2025. Google’s approach shares similarities with Apple’s Sensitive Content Warning feature available on iOS and macOS, indicating a broader industry direction towards providing users with tools to manage potentially unwanted explicit content directly on their devices.
Source: Winbuzzer / Digpu NewsTex