Introduction
Global technology company Google officially announced on Wednesday the launch and activation of a revolutionary tool for computer use and direct screen control, making this advanced tool a built-in and essential part of the updated and latest version of its advanced artificial intelligence model named Gemini 3.5 Flash. This advanced and bold technical step is primarily aimed at enabling the developer and engineering community to build and design autonomous AI agents that possess the actual and practical ability to clearly see what is displayed on users’ screens, analyze visual data and accurately deduce information from it, and interact actively and directly with complex user interfaces across multiple computing environments. These environments include web browsers, mobile devices, and traditional desktop operating systems, representing an unprecedented qualitative leap in integrating autonomous agent capabilities into a single flexible model that supports developer innovations and facilitates complex administrative workflow automation.
- From a standalone computing model to a comprehensive built-in tool
- Built-in development tools via the Gemini API
- Product management statements and enterprise automation enhancement
- Strict security guarantees to protect live working environments
- The evolution context of the Gemini model within Google’s overarching strategy
From a standalone computing model to a comprehensive built-in tool
This powerful and highly precise interactive capability was previously available to developers exclusively through the standalone and entirely separate computer-using model known as Gemini 2.5, but today it has become natively, fully, and directly embedded in the core Flash model that many applications rely on. This deep structural integration practically means that developers and programmers no longer urgently and continuously need to call a separate model dedicated exclusively to building agents that interact smoothly with complex graphical user interfaces. Instead of this complexity, the computer use and screen control feature can be activated simply and flexibly as one of several built-in tools directly available to users.
Built-in development tools via the Gemini API
This tool operates with high efficiency within the comprehensive Flash system, performing its tasks alongside other essential and powerful tools such as code execution, complex database searches, and calling required programming functions. All of this is executed with high efficiency through the Gemini API, as well as the specialized Gemini Enterprise Agent platform designed to serve the business sector, formerly known in tech circles and among developers as Vertex AI.
Product management statements and enterprise automation enhancement
In this prominent technological context, Google Product Manager Mathieu Quirion described this precise integration process as an important and pivotal step that gives the Flash model the actual and practical ability to “clearly see screens, think and analyze the displayed logical content, and take necessary and appropriate actions automatically and quickly,” according to documented statements reported by the news network The Next Web. Google added in its accompanying technical clarifications that this new and important update directly and noticeably contributes to improving the model’s overall performance, particularly in executing long-term tasks and complex enterprise automation processes, such as continuous and periodic software testing and the completion of complex cognitive tasks within specialized professional applications without the need for constant human intervention.
Strict security guarantees to protect live working environments
To address the significant security risks that may arise from allowing autonomous AI agents to operate freely and independently in live and open corporate computer environments, Google introduced special competitive training targeted at countering adversarial attacks and ensuring that attackers cannot exploit malicious prompt injection vulnerabilities. Furthermore, the established company offers two additional options for strict enterprise and security guarantees: the first is requiring explicit and direct confirmation and approval from a human user before the agent executes highly sensitive or irreversible actions, while the second option automatically, immediately, and decisively stops the smart agent if the system detects any indirect or suspicious attempt to hijack prompted instructions. To ensure a secure and stable working environment for developers, the company also provides a hosted and protected preview and testing environment in close collaboration with the Browserbase platform to test these advanced capabilities with complete safety.
The evolution context of the Gemini model within Google’s overarching strategy
Gemini 3.5 Flash was first released at Google’s annual developer conference held on May 19, 2026, as the company’s fastest interactive and agentic AI model to date, offering frontier-level performance for autonomous agents and complex programming tasks. The model was launched at the time equipped with a massive and impressive one-million-token input context window, full native support for function calling, precise structured outputs, and code execution, but it did not initially include the computer use and screen control feature. This latest update, documented and announced as a public preview version, aims to completely bridge this strategic gap, positioning Gemini 3.5 Flash as a single, comprehensive model capable of seamless and intelligent handling across the full range of multiple AI agent tasks, saving developers the trouble and cost of coordinating between multiple separate models.
Frequently Asked Questions
Question: What new tool was recently added to the Gemini 3.5 Flash model?
Answer: Google added direct screen control and computer use tools as a built-in, essential feature in the model to enable intelligent agents to interact with graphical interfaces.
Question: How will this new technical integration affect developers and their daily work?
Answer: It will spare developers the need to call and run a separate, independent model, facilitating the process of building autonomous agents, saving time, and reducing operational costs.
Question: What are the security guarantees and measures put in place by Google to protect systems and enterprises?
Answer: Google provided specialized training against malicious prompt injection attacks, alongside an option to request human approval before executing sensitive actions, and immediate automatic termination upon sensing a breach.
Question: When was the initial release of the Gemini 3.5 Flash model launched to the public?
Answer: The model was first unveiled and launched during Google’s annual developer conference held on May 19, 2026.
