The Prompt API is now on by default in Chrome (adsm.dev)

Google has officially integrated the Prompt API into Chrome 148, allowing developers to leverage on-device language models for web applications by default. This marks a significant move from earlier beta versions and promises enhanced interactivity as web pages can now engage directly with AI models like Gemini Nano. These models are downloaded to the user's device after initial interaction, with either a GPU build for higher performance or a CPU build for less powerful systems, a decision based on the hardware capabilities detected at launch. The implications for the AI/ML community are substantial. Running models directly on user devices can reduce latency and enhance privacy, but it also raises concerns about resource consumption, automatic model downloads, and potential misuse in web contexts. Critics, including developers from Mozilla and WebKit, have voiced concerns over user consent and management of system resources, questioning the API's transparency and reliability. As the Prompt API develops, it could reshape how web applications implement AI, while also igniting discussions about responsible AI deployment and user control in browser environments.