Google Quietly Buying Play Store Code to Train AI Coding Tools

Google has been contacting Android developers with an offer to buy access to their app codebases to improve its AI coding tools, according to a 404 Media report. The program, described as a "confidential content offer pilot," targets developers with apps on the Play Store — including one with millions of downloads who shared the email anonymously.
Key Details from the Email
- What Google wants: access to production code, archived projects, prototypes, and side projects — including "the active production codebase powering your current app" and "archives of prototypes and side projects no longer in use."
- License: non-exclusive, developer retains 100% IP and can monetize code elsewhere.
- Payment: developers get paid for sharing code, but no specific amounts were disclosed.
- AI connection: the email does not mention AI, but a linked page references "partnerships to improve our AI products" and training on "non-public content."
The email frames the program as a way for developers to generate extra revenue while helping improve Google's developer tools. "Whether it's the active production codebase powering your current app, or archives of prototypes and side projects no longer in use, that code could have untapped value," the email states.
Why This Matters
Google is behind Anthropic's Claude Code and Microsoft's Copilot in code generation AI. This program suggests Google's web-scraped data isn't sufficient for competitive coding AI, echoing its $60 million Reddit data deal (with mixed results). As the article notes, "the fact that Google is trying to buy code from developers suggests that the company hasn’t been able to create a good enough coding AI using content that it can scrape from the web."
For developers, the program raises questions about code quality requirements, pricing, and the long-term impact on Google's AI tools. The confidentiality clause indicates Google wants to avoid public scrutiny during the pilot.
📖 Read the full source: HN AI Agents
👀 See Also

Gemini 3.1 Flash Live: Google's latest audio model with improved benchmarks and watermarking
Google released Gemini 3.1 Flash Live, an audio model scoring 90.8% on ComplexFuncBench Audio and 36.1% on Scale AI's Audio MultiChallenge. It's available via Gemini Live API in Google AI Studio and includes SynthID watermarking.

AI Vendor Lock-In Escalates: Switching Models Now Costs More Than Most Expected
A Zapier survey of 542 US executives shows 90% thought they could switch AI vendors in under 4 weeks, but 58% of actual migrations failed or took far longer. Meanwhile, OpenAI raised GPT-5.2 input token pricing from $1.25 to $5.75, and Anthropic moved Claude enterprise to dynamic pricing, potentially doubling or tripling costs for heavy users.

OpenAI Training Costs Projected to Exceed Anthropic's by 4-5 Times Annually
According to confidential financials reported by the Wall Street Journal, OpenAI expects to spend 4-5 times more on training than Anthropic each year for the next five years. The expense scale is described as mind-boggling.

RTX 4090 vs H100 for Fine-Tuning Llama-3-8B: A Cost-Performance Comparison
A developer tested fine-tuning Llama-3-8B on both an RTX 4090 and rented H100 instances. The 4090 setup cost $2,000 upfront and took 24 hours, while H100 rental cost about $80 and completed in 4 hours.