Google has added built-in computer use capabilities to Gemini 3.5 Flash, allowing developers to create AI agents that can navigate websites, perform tasks and interact with digital environments without needing a separate specialized model.
Available through the Gemini API and Gemini Enterprise Agent Platform, the feature enables Gemini to complete actions such as searching flight websites, comparing prices and returning results autonomously. Google also demonstrated the model’s ability to play games like 2048 by making decisions on its own.
To address security concerns, Google has introduced safeguards that can require user approval before sensitive actions and automatically stop tasks if a prompt-injection attack is detected. The company also recommends using secure sandboxes, strict access controls and human oversight when deploying AI agents. The new computer use feature is available starting today.
Source: Android Authority
