Google has asserted in court that it retains the ability to train its search-specific AI models, including AI Overviews, using publicly available web content, irrespective of publishers opting out of Google's broader AI training programmes. This revelation, disclosed by a Google VP of Product, highlights a distinction in how Google approaches content usage for general AI development versus search-focused applications.
The core of the issue revolves around the interpretation of content opt-outs. While publishers can signal their desire to exclude their content from being used in training Google's general AI models, this restriction does not automatically extend to AI models specifically designed to enhance Google's search functionalities. Google's rationale likely hinges on the argument that indexing and processing web content for search is a fundamental aspect of its core service, and that AI-driven improvements to search fall under this umbrella.
The implications of this policy are significant for content creators and publishers. It suggests a limited control over how their content is utilised by Google, particularly in the context of AI-powered search enhancements. As AI Overviews and similar features become increasingly integrated into the search experience, the ability of Google to train these systems on a wide range of web content, even against the explicit wishes of some publishers, could further consolidate Google's dominance in search and AI.