Google is reaching out directly to Android app developers with an unusual proposition: give us your proprietary codebases, and we'll pay you for it. According to a confidential email obtained by 404 Media's Jason Koebler, the tech giant has started recruiting Play Store creators to license their private source code for training its Gemini AI models and improving developer tools.

The Confidential Offer

The email frames this as a 'mission-driven opportunity' and offers developers several incentives: additional revenue for sharing code powering their apps, early-adopter status as pilot partners, and the ability to retain full intellectual property rights through a non-exclusive license. Developers keep 100% of their IP, their apps remain entirely theirs, and they can monetize the same data elsewhere. The pitch emphasizes real-world production-tested code as valuable for 'understanding complex logic' and developing 'coding evals and benchmarks.' One anonymous developer described it as 'a unique occasion to help transform tools and products, support the developer ecosystem, and unlock new revenue.'

Why Google Needs This Data

The timing isn't coincidental. Google's Gemini has been playing catch-up against GitHub Copilot, which remains the gold standard for real-time code autocomplete with seamless IDE integration and lightning-fast boilerplate generation. Meanwhile, Anthropic's Claude Code specializes in agentic coding workflowsβ€”executing complex tasks across entire codebases. The tool became so effective that Microsoft recently started revoking employee licenses after engineers flocked to it en masse during late 2025.

I/O 2026 Response

Google's clearly aware of the gap. At this year's I/O conference, the company unveiled Gemini 3.5 Flash and Pro alongside Antigravity 2.0β€”its direct answer to Claude Code designed to help developers orchestrate teams of autonomous coding agents. But raw model announcements mean nothing without quality training data, which explains why Google is now willing to pay for access to private repositories rather than relying solely on publicly available code.

A New Playbook

Paying developers for proprietary codebase access represents a fresh tactical shift in the AI training arms race. Microsoft experimented with something similar last year through its Publisher Content Marketplace, allowing publishers to set licensing terms and receive payments when AI companies used their content. Google appears to be adapting this model specifically for codeβ€”a more defensible approach than scraping public repositories without consent.

Key Takeaways

  • Google's non-exclusive licensing means developers keep IP rights while earning revenue from previously untapped assets
  • The move signals that freely available open-source code isn't sufficient for cutting-edge AI training anymore
  • Microsoft's Claude Code popularity problem shows why even internal companies can't resist superior tools

The Bottom Line

Google paying for private codebases is a tellβ€”it means they've exhausted the easy data sources and recognized that quality training data has become a competitive moat. If you're an Android developer with battle-tested production code sitting in private repos, congratulations: your legacy codebase just became an asset.