Why did Google silently install a 4GB AI model?

Chrome controversy: a 4GB Gemini model installed without consent

Google faced backlash after users reported that Google Chrome had installed a 4GB AI model onto their devices as part of Gemini-based capabilities. The core issue is not only that a large download occurred, but that it happened in a way that raised questions about user consent, transparency, and control.

Coverage emphasized several risk points:

• Consent and expectations: Users said the model was installed without clear, meaningful permission.
• Bandwidth and cost impact: A 4GB download can be significant for mobile or metered connections, and the incident triggered complaints about practical real-world costs.
• Environmental concerns: Some reporting framed the incident as increasing energy or resource usage due to automatic model downloads.

What users could do

Follow-up reporting indicated there were ways to manage or disable the behavior. In one case, instructions circulated for removing or disabling Gemini in Chrome.

Why it matters

A large model download inside a mainstream browser changes how “AI features” are experienced: instead of a lightweight server call, users may be pulling substantial data to their devices. That shifts the trust bar from “does the feature work?” to “do you clearly control when data is downloaded and used?”

The episode also underscores an emerging tension in AI product rollouts: speed of deployment versus user agency. When the browser is the distribution channel, broad adoption makes even small implementation details large in scale.

In short, the controversy centered on Chrome getting a hefty AI payload on users’ devices in a way that didn’t align with what many users expected for consent and transparency.