A solo indie developer behind the tower defense game Operation Octo on Steam has reported that Google's AI accessed and disclosed unrevealed information about their project. The developer, known as Klub Kofta Studio, became aware of the issue after a user on their Discord server queried a Google AI about Operation Octo and received obscure details that had not been made public.
Klub Kofta shared their experience on Reddit, explaining that upon their own inquiry to the AI about future content for their game, the AI produced a highly specific character name, 'Vantage Tripod.' This name had not been shared publicly, and the developer stated that the only place this information existed digitally was within a private Google Doc. The developer emphasized that they had not directly interacted with the AI in a way that would have provided this information.
Despite Google's previous assurances that information within Google Drive or other Google services is not used to train its AI models without explicit permission, a shift in policy suggests a possible explanation. While a Google Product VP stated in 2023 that Workspace data was not used to train AI outside of Workspace, this statement appears to be outdated.
TechCrunch reported in July that Google announced via a customer email in June that it now uses saved history and media to "provide, develop, and improve its services (such as training generative AI models) and to protect Google, its users, and the public with the help of human reviewers." This change means that user data, including content from private documents if associated with a Google account, could potentially be used for AI training. Crucially, this data usage is opt-out rather than opt-in, meaning users may have already consented to sharing their information by default if they use Google services. A direct link to Google Personalization settings is available for users wishing to review or disable this feature.