Google’s AI‑tool divide

Google is reportedly letting some engineers use stronger coding AI tools while keeping most teams on its own systems. (timesofindia.indiatimes.com) The Times of India, citing Business Insider, reported on April 22 that some employees in Google DeepMind can use Anthropic’s Claude for coding, while most Google engineers are limited to internal tools such as Gemini. (timesofindia.indiatimes.com) The report said teams are being sorted by function and risk tolerance, and some engineers have tied AI-use goals to performance reviews. It also said some employees see Claude as more effective for coding than Google’s in-house tools. (timesofindia.indiatimes.com) Google had already moved in the opposite direction for the broader engineering organization. In June 2025, the company sent software engineers formal guidance that said everyone was expected to use AI to boost productivity and stressed code review, security, and maintenance. (9to5google.com) That guidance followed a public push from Sundar Pichai, who said in April 2025 that more than 30% of Google’s code was being generated by AI before human review and that AI had raised engineering velocity by about 10%. (9to5google.com) Inside Google, the split lands as the company is trying to prove that its own AI stack can serve both employees and customers. On Alphabet’s February 4, 2026 earnings call, Pichai said Google had sold more than 8 million paid seats of Gemini Enterprise and that the Gemini app had more than 750 million monthly active users. (blog.google) The wider software industry has moved toward near-universal AI use, but not universal trust. Google Cloud’s 2025 DORA report, based on nearly 5,000 technology professionals, found 90% were using AI at work, more than 80% said it improved productivity, and 30% said they trusted it only “a little” or “not at all.” (blog.google) Google is also facing outside pressure in coding tools. Bloomberg reported on April 21 that leaders were worried Google was trailing rivals such as Anthropic in AI coding and were trying to pull scattered internal efforts under one banner. (bloomberg.com) The company’s case for limiting outside tools is familiar: security, internal infrastructure, and “dogfooding,” the practice of making employees use Google products to test and improve them before release. The friction comes when one part of the company gets exceptions and another does not. (timesofindia.indiatimes.com) Google DeepMind chief executive Demis Hassabis pushed back on criticism that Google’s AI adoption was lagging, calling one circulating claim “completely false” and “pure clickbait.” The new reporting suggests the argument inside Google is no longer about whether engineers should use AI, but which AI they are allowed to use. (timesofindia.indiatimes.com)