Google is currently testing a new API designed to leverage machine learning models for real-time language translation. This tool aims to make it easier to translate web pages and process inputted text instantly.
According to a report by Bleeping Computer, this feature is being developed by Chrome’s AI team to enhance the browser’s existing translation capabilities while allowing users to download additional language models when needed.
Although Chrome and Microsoft Edge already offer built-in translation tools, these features often struggle with complex or interactive web pages. For example, Chrome may miss certain sections of a dynamically updated site. To overcome this limitation, many websites deploy their own translation systems, which can be resource-heavy and slower than browser-native options. Google’s new API intends to simplify this by giving developers direct access to Chrome’s AI-powered translation engine via a straightforward JavaScript API.
Currently, websites either rely on cloud-based translation services or bring their own machine learning models, which run on technologies like WebAssembly. Google’s proposed API would allow websites to harness the browser’s built-in translation capabilities instead, reducing the demand for external resources. If Chrome doesn’t have a specific language model, the feature will automatically download one to provide accurate translations for both web content and user input in real-time.
However, privacy concerns have been raised about this new system. Google has acknowledged that the API could expose identifiable information, such as a list of supported languages and their availability status. This could potentially allow websites to “fingerprint” users based on their language settings. To mitigate this risk, Google is considering grouping languages or limiting the exposure of language data based on the user’s region.
While there is no confirmed release date, Google is actively working with the open-source community to refine this feature, ensuring it balances performance, privacy, and user convenience.
