Google AI-tool divide

A split has opened inside Google over which artificial intelligence tools engineers can use to write code: Anthropic’s Claude for some teams, Google’s own systems for others. (businessinsider.com) Business Insider reported on April 21 that some Google DeepMind employees had been given access in recent months to Claude for coding, while many other Google engineers were limited to internal Gemini-based tools. A Google spokesperson did not respond to Business Insider’s request for comment. (businessinsider.com) The divide sits inside a broader push by Google to get employees to use more artificial intelligence at work. Business Insider reported that Google generally bars staff from using outside tools unless they are Google-built or customized for internal use, and that some engineers were frustrated colleagues had exceptions. (businessinsider.com) The dispute is centered on coding assistants, which are chatbots trained to suggest, rewrite, and test software. Bloomberg reported on April 21 that Google leaders were worried about falling behind Anthropic in tools for software development, especially in products sold to businesses. (bloomberg.com) Google’s answer has included an internal effort to improve its own coding models. The Information reported on April 20 that Google assembled a strike team of researchers and engineers to sharpen its artificial intelligence coding systems as it tries to automate more of its own software work. (theinformation.com) That effort has drawn attention from the top. The Verge, citing The Information, reported that Sergey Brin told DeepMind staff in a memo that “every Gemini engineer” should use internal agents for complex multistep work as Google tries to close the gap with Anthropic on coding tools. (theverge.com) Google has reasons to keep engineers on in-house systems beyond competition. Business Insider reported that much of Google’s internal infrastructure is custom-built, and that the company sees “dogfooding” — employees using the same products the company ships — as a way to improve those tools faster. (businessinsider.com) Other companies have taken a looser approach. Business Insider reported that Meta employees can use Claude internally, underscoring how large technology companies are making different trade-offs between security controls, product loyalty, and access to the strongest outside models. (businessinsider.com) Inside Google, the immediate question is not whether employees will use artificial intelligence for coding, but whose artificial intelligence they will be expected to use. (businessinsider.com)