A new internal divide has emerged at Google regarding access to external artificial intelligence tools, creating tension between employees at Google DeepMind and those in other divisions. While the company generally prohibits staff from using tools not made by Google or tailored for internal use, some DeepMind engineers have recently been granted permission to use Anthropic's Claude for coding tasks. This exception has frustrated engineers elsewhere in the organization who are restricted to using Google's internal Gemini models, with some reporting that Gemini is currently inferior for coding purposes. The friction is heightened by Google's broader push for AI adoption across the company. Many engineers are now assigned specific AI integration goals that factor into performance reviews, with some expected to use AI to generate code or build efficiency tools. The disparity in tool access contrasts with the policies of other major tech firms; for instance, Meta employees are permitted to use Claude internally. Google maintains its strict internal policy to ensure its custom-built infrastructure is utilized effectively and to encourage dogfooding, a practice where employees test and refine products they plan to launch. The issue came to public attention following comments by Steve Yegge, a computer programmer and blogger, who claimed on the social media platform X that Google's internal AI adoption was lagging significantly. He compared Google's engineering AI footprint to that of a traditional tractor manufacturer, John Deere. In response, Demis Hassabis, CEO of Google DeepMind, issued a sharp rebuke, calling the claims false and mere clickbait. Hassabis urged Yegge to focus on his own work rather than spreading what he termed nonsense. Yegge subsequently replied, stating he had received corroboration from Google employees supporting his claims. He reported that when the topic of equalizing access to Claude was discussed internally, some DeepMind engineers reacted strongly. According to Yegge, the proposed solution was to remove Claude access for everyone, a move that DeepMind reportedly opposed so vehemently that several engineers threatened to leave the company if the restriction remained. The situation highlights the ongoing challenges of balancing proprietary technology strategies with the rapid advancement of external AI models in the workplace. A Google spokesperson declined to comment on the matter.