Google splits on internal AI use
- Reports describe an internal divide at Google over which models engineers are allowed to use, with some using Claude and others pushed to Gemini. - Google is also embedding Gemini-powered 'auto browse' features into Chrome and adding Workspace Intelligence with admin privacy controls. - Those developments show enterprise AI moving into everyday workflows while elevating governance and data-access questions ( ).
A split has opened inside Google over which artificial intelligence tools employees can use, even as the company pushes Gemini deeper into work software. (timesofindia.indiatimes.com) A Times of India report, citing The Information, said some engineers at Google DeepMind have access to Anthropic’s Claude for coding tasks, while most Google engineers are required to use internal tools including Gemini. Google had already told software engineers in 2024 that they needed approval before using external artificial intelligence tools. (timesofindia.indiatimes.com (timesofindia.indiatimes.com)) On April 22, Google also said it would add Gemini-powered “auto browse” features to Chrome Enterprise, letting workers automate web tasks such as research and data entry inside the browser. TechCrunch reported the announcement was part of Google Cloud Next. (techcrunch.com) Google made a parallel move in Workspace the same day with Workspace Intelligence, which it described as a system that gives Gemini a real-time view of a user’s work across Gmail, Chat, Calendar, Drive, Docs, Sheets and Slides. Google said administrators can choose which data sources the system may use in the Admin console. (workspaceupdates.googleblog.com) Google’s admin documentation says Gemini Business and Enterprise can connect to Workspace data including Gmail, Drive and Calendar, and that organizations can decide whether Gemini may access that data. Google’s help pages also say the default setting for Gemini features in Workspace services is on. (support.google.com 1) (support.google.com 2)) In practice, Google is putting the same assistant into two layers of office work at once: the browser where employees use web apps, and the productivity suite where company files and messages live. Chrome’s new features follow earlier Gemini additions to Chrome, including saved “Skills” prompts announced on April 14. (techcrunch.com 1) (techcrunch.com 2) Google is also widening Gemini’s reach inside Workspace itself. On April 22, Google announced AI Overviews in Gmail search, general availability for AI Overviews in Drive, new Gemini features in Docs, and Gemini-built spreadsheets in Sheets, each tied to Workspace Intelligence. (workspaceupdates.googleblog.com 1) (workspaceupdates.googleblog.com 2) (workspaceupdates.googleblog.com 3) (workspaceupdates.googleblog.com 4) Those rollouts put Google in an awkward position if internal teams believe rival models work better for some jobs. The same week Google expanded Gemini across Chrome and Workspace, reports said some of its own engineers were still using Claude for coding. (techcrunch.com) (timesofindia.indiatimes.com) Google’s public line on the Workspace side is that permissions and privacy controls stay in place. The company said Workspace Intelligence respects existing user access controls, and Help Net Security reported that Google says customer data is not used to train generative artificial intelligence models outside the service or for advertising. (workspaceupdates.googleblog.com) (helpnetsecurity.com) The next test is whether companies accept that trade: broader Gemini access in exchange for tighter admin rules over what the assistant can read. Inside Google, that question already appears less settled than the product launch suggests. (support.google.com) (timesofindia.indiatimes.com)