Google AI Leaked Solo Dev’s Unreleased Game Content From Private Docs

A solo indie developer discovered that Google’s artificial intelligence had accessed and publicly shared unrevealed information about their game—information that existed only in a private Google Document. The incident raises significant questions about how Google’s AI systems interact with user data and what protections developers have when storing sensitive content on the platform.

Klub Kofta Studio, the solo developer behind the tower defense game Operation Octo on Steam, first learned of the breach through an indirect route. A user in the developer’s Discord server had been asking Google’s AI about the game and reported receiving obscure details that should have been inaccessible. When Klub Kofta decided to test this directly, they asked the AI about unreleased future content—and received an exact answer: the character name “Vantage Tripod,” a detail never shared publicly anywhere.

“Out of curiosity, I chimed in and asked the player to ask the AI about future contents in my game,” Klub Kofta explained in their Reddit post. “Somehow, the AI spitted out the exact & highly specific character name ‘Vantage Tripod,’ which I had never mentioned to anyone. As far as I know, the only place where the info exists in a digital form is inside one of my own Google Docs, and this was NOT me speaking to the AI!”

The discovery would be confusing enough on its own. But it became alarming when paired with Google’s official public stance on data privacy and AI training.

Google’s Public Promise on Data Privacy

For years, Google has maintained that user data stored in its Workspace services remains strictly protected from AI training. In 2023, Yulie Kwon Kim, Google’s Product Vice President, published an explicit statement promising that the company does “not use your Workspace data to train or improve the underlying generative AI and large language models that power Bard, Search, and other systems outside of Workspace without permission.”

That statement represented a direct reassurance to millions of users who rely on Google Drive, Google Docs, and other Workspace services to store confidential business information, creative projects, and unpublished content. For developers like Klub Kofta, storing game plans, character designs, and story concepts in Google Docs seemed safe—protected by explicit corporate policy.

But that public reassurance masked a significant gap between Google’s announced policy and what the company actually disclosed to users.

The Hidden Policy Shift

Read also: Modder Resurrects Six Years of Deleted Destiny 2 Locations for Free Exploration · GameCenter CX Debates Wii and PS3 as Official Retro Systems · Machine Party: Buckshot Roulette Creator’s Deadly Party Game · Quake Celebrates 30 Years as NIN Merch Sells Out Instantly

Google did announce changes to how it uses user data for AI training—the announcement just didn’t come through the channels most users monitor. Instead, as TechCrunch reported in July, Google sent the news via customer email in June. That email contained a policy update stating that Google 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.”

What Actually Changed

This marked a departure from the 2023 statement. Under the new policy, content stored in Google’s ecosystem could now feed directly into the training of AI models—a shift with major implications for anyone storing sensitive information on the platform. The scope extended beyond just generative AI systems, as Google indicated that human reviewers would also have access to this content as part of the training and improvement process.

How Users Were Enrolled

For most users, the practical impact of this policy change came down to a single word: opt-out. Google didn’t require users to actively agree to having their data used for AI training. Instead, Google made data-sharing the default setting, forcing users to navigate to Google’s Personalization settings and manually disable the feature if they wanted to maintain the previous level of privacy.

The Opt-Out Trap

This distinction matters enormously. Opt-in systems require explicit user consent before data sharing begins. Opt-out systems assume consent unless users take action to refuse. By choosing an opt-out model, Google ensured that millions of users would automatically have their private documents, emails, and media files enrolled in AI training whether they knew about the policy change or not.

For a solo developer like Klub Kofta, it’s unlikely they saw or processed a customer email about a June policy update—especially one that fundamentally altered how their unpublished work could be accessed and used. The result: Operation Octo’s unreleased character concepts ended up in Google’s AI training data without explicit consent.

The Developer’s Dilemma

Developers regularly use cloud services to store works-in-progress, design documents, narrative outlines, and creative assets. These materials represent unreleased intellectual property—exactly the kind of content that should remain confidential until publication. When a developer stores a character name exclusively in a private Google Doc, they reasonably expect that information to stay private. Google’s previous public statements reinforced that expectation. But the June policy change meant that assumption was no longer valid for anyone who hadn’t specifically opted out.

Implications for Game Development and Beyond

The implications extend beyond individual developers. Game studios of all sizes rely on cloud storage for collaborative work, and many probably store confidential information in Google Workspace. The policy change means all that content is now fair game for AI training unless the studio actively discovers and disables the feature.

The incident also raises fundamental questions about consent, corporate transparency, and how AI systems should be trained. When a developer’s secret character concept appears in an AI’s responses, where does responsibility lie? With the developer for not opting out of a hidden policy? With Google for making opt-out the default? With the AI user for querying it about a game?

What Game Developers Must Know

For anyone storing sensitive information in Google’s ecosystem, the lesson is clear: assume nothing is off-limits for AI training unless you’ve actively disabled the feature. The default setting no longer protects unpublished work, unreleased projects, or confidential business information. Operation Octo’s leaked character name is a small consequence compared to what could happen if larger creative studios or companies with trade secrets stored sensitive information in Google Workspace under the old assumptions.