Google is quietly paying Android developers for access to their app source code — including abandoned prototypes and archived side projects — to fuel its AI model training, according to a report by 404 Media.
The program, framed internally as a "confidential content offer pilot," targets a select group of Google Play developers with an email from the Google Partnerships team.
The pitch positions it as an easy revenue opportunity: sell your codebase (the full working source code behind an app), and Google will put it to work. What the email conspicuously omits is any mention of artificial intelligence — though a link buried in the message leads directly to a page about "partnerships to improve our AI products."
It acknowledges that Google is now actively paying for non-public content beyond what it can scrape freely from the web, calling it a chance to create "mutually beneficial collaborations." Developers who participate retain 100% of their intellectual property rights under a non-exclusive license, meaning they can still monetize or publish their code elsewhere.
The significance here goes beyond one company's data shopping. Most AI training data is sourced from public content scraped across the internet — usually without compensation to creators. Android app code, by contrast, is inherently private.
Google's willingness to pay for it signals that the industry's freely available training data pool may be running dry. The company paid Reddit $60 million for a similar arrangement back in 2024, with uneven results.
Anthropic's Claude Code has surged in developer adoption, and Microsoft's GitHub Copilot remains deeply embedded across enterprise workflows.
Google's Gemini-based coding tools have struggled to keep pace, and buying real-world, production-tested Android codebases could help close that gap — particularly for understanding complex application logic and building coding benchmarks (standardized tests that measure how well an AI model writes or completes code).
For developers receiving the email, the decision is nuanced. The IP protections appear solid on paper, but handing proprietary production code to a major platform partner carries its own risks — particularly for developers whose apps compete with Google's own ecosystem products.
Google has not publicly commented on the program.