Google forcing employees to use internal AI coding tools that they do not trust

There is a quiet but growing friction inside Google, and it revolves around something the company has been aggressively pushing - AI. While leadership wants employees to rely more on AI tools for coding and productivity, not everyone inside the company is convinced the current approach is working or fair. At the centre of the issue is access. A section of engineers working under Google DeepMind has reportedly been allowed to use Claude, one of the most widely used AI coding assistants in the industry today. But for most other teams across Google, that option is off the table. They are expected to rely on the company’s own models, primarily Gemini.

AI push meets internal resistance at Google

This uneven access has not gone unnoticed. According to Business Insider, engineers in other divisions are said to be frustrated, especially as some believe that Google’s internal tools are still catching up when it comes to coding performance. The result is a strange situation where employees are being pushed to use AI more in their daily work, but many feel they are not being given the best tools available to do so.

The pressure is not just informal. In some teams, AI usage is now becoming part of performance evaluations. Engineers have reportedly been given clear targets, not only to use AI tools while writing code but also to build systems that improve efficiency using AI. For a company that has long prided itself on engineering excellence, this is a change in how productivity is being measured.

Google, however, has its reasons for keeping things in-house. Much of its internal infrastructure is deeply customised, which makes integration with external tools more complicated. There is also a long-standing belief in “dogfooding” — the idea that employees should use the same products that are being built for users. In theory, this helps improve the tools faster. In practice, some employees feel it is limiting their ability to work efficiently.

The contrast becomes sharper when compared to competitors like Meta, where employees reportedly have more flexibility to use external AI tools, including Claude. That openness has raised questions among some Googlers about whether the company’s strict approach is slowing things down internally.

The debate spilled into public view recently after programmer Steve Yegge claimed that AI adoption inside Google was lagging. He wrote that the company’s engineering teams were not using AI at the level many might expect, drawing an unflattering comparison to traditional industries.

That comment triggered a sharp response from Demis Hassabis, who dismissed the claims outright, calling them "completely false" and "pure clickbait." But the conversation did not end there. Yegge later said he had heard from Google employees who supported parts of his claim, including the uneven access to tools like Claude.

Perhaps the most telling detail is what reportedly happened when the idea of equal access came up internally. Instead of expanding Claude access to more teams, one proposal was to remove it entirely — even from DeepMind. That suggestion did not go down well. According to reports, some engineers pushed back strongly, with a few even threatening to leave if access was taken away.