Gemini 3.5 Flash: Integrated Computer Usage

A Major Advancement with Gemini 3.5 Flash

The integration of computer usage in Gemini 3.5 Flash marks a significant milestone in the development of digital agents. This feature now allows for the creation of agents capable of interacting seamlessly across a multitude of platforms, whether they are browsers, mobile devices, or desktops.

Previously, computer usage was a distinct feature, available only in Gemini 2.5. With this new integration, Gemini Flash becomes a centralized tool, optimizing agent performance for complex tasks. Gemini already excels in calling functions and utilizing built-in tools such as search and mapping. Developers can now design custom agents that not only see and reason but also act in various digital environments. This paves the way for advanced applications in software testing automation and knowledge management within enterprises.

Easier Access for Developers and Businesses

Developers and businesses now have the opportunity to leverage this feature through the Gemini API and the Gemini Enterprise Agent Platform. This integration simplifies the process of creating sophisticated agents capable of analyzing and categorizing the features of Gemini applications. Furthermore, 3.5 Flash is capable of auditing its own documentation, thereby identifying potential accessibility issues.

Enhanced Security for Risk-Free Use

Security is a priority in the use of computers with Gemini 3.5 Flash. To minimize risks associated with prompt injections, targeted adversarial training has been implemented. Additionally, two optional backup systems are offered to businesses. These systems allow for explicit user confirmation for critical actions and automatically halt tasks upon detecting an indirect prompt injection.

This "defense in depth" approach is complemented by recommendations for using secure sandboxing, human verification, and stringent access controls. Developers are encouraged to consult the documentation on best practices to enhance the security of their applications.

Initial Feedback and Implementation

Early users of this new feature are already reporting significant benefits. For those looking to explore these capabilities, Browserbase offers a demonstration environment to test the functionalities. Developers can also access our detailed documentation and reference implementation via the Gemini API and the Gemini Enterprise Agent Platform.