Gemini 3.5 Flash: AI Capable of Operating a Computer Independently
Google continues to innovate with their artificial intelligence (AI) technology, and one of the latest breakthroughs is Gemini 3.5 Flash. In this version, the AI not only provides recommendations but also has the ability to operate a computer autonomously. This feature brings significant changes for users, both individually and in business contexts.
Integration with Workspace Applications
Alongside the development of Gemini, Google is increasingly integrating this AI with workspace applications such as Google Drive. The goal is to make Gemini more useful for consumers. Additionally, Google is focusing on developing new features aimed at enterprises and developers, including ease in creating AI agents that can think, navigate, and take actions in various environments.
Introduction of New Features in Gemini 3.5 Flash
One of the new features introduced is the computer usage capability, now integrated as a built-in tool within Gemini 3.5 Flash. Previously, developing custom AI agents required a separate Gemini 2.5 computer usage model, but this is no longer necessary. The new model can be accessed by developers through the Gemini API or the Gemini Enterprise Agent Platform.
To demonstrate its capabilities, Google has created a Browserbase instance. In this context, users can give commands to the model to perform specific tasks. For example, when I asked Gemini to find the cheapest flight from New Delhi to Tokyo, the AI was able to visit three separate ticket booking sites, enter departure and return dates, and search for available tickets. The result was that it provided the best options for me to choose from.
Security and User Control
Of course, the ability to control a computer and perform tasks automatically raises questions about security, especially for users in the enterprise sector. To address these risks, Google has implemented targeted adversarial training for this model. Additionally, two new security measures have been introduced for computer usage with Gemini 3.5 Flash. First, the model can be set to require explicit user confirmation before performing sensitive or irreversible actions. Second, the AI can automatically stop tasks if it detects prompt injection attacks.
Google also recommends that developers combine these security measures with secure sandboxes, strict access controls, and "human-in-the-loop" verification to enhance the safety of using this AI.
Accessing Gemini 3.5 Flash
Computer usage with Gemini 3.5 Flash is now available. This is a major step in advancing AI to a higher level, providing convenience and efficiency for users. With increasingly sophisticated capabilities, Gemini 3.5 Flash promises a new experience in interacting with technology.
From the ability to perform searches to playing games, Gemini 3.5 Flash demonstrates remarkable potential. Over time, we can expect to see more practical applications of this technology across various fields.
Source: https://www.androidauthority.com/gemini-3-5-flash-computer-use-model-3681381/
