Google has announced that computer use is now integrated directly into Gemini 3.5 Flash, marking a major step forward in its Gemini AI platform. This development opens the door for developers and enterprises to leverage agentic computer use tasks with the model's highest performance to date. Previously, computer use capabilities were limited to a standalone Gemini 2.5 computer use model, but now they are available natively in the latest Gemini Flash version. The rollout is available to developers and enterprise users via the Gemini API and Gemini Enterprise Agent Platform, targeting organizations seeking advanced automation solutions.
Gemini 3.5 Flash's built-in computer use enables custom agents to observe, reason, and act across browser, mobile, and desktop environments. This supports complex use cases such as continuous software testing and enterprise knowledge work. Technical upgrades include targeted adversarial training to address prompt injection risks and two optional enterprise safeguard systems:
- Requiring explicit user confirmation for sensitive actions.
- Halting tasks if indirect prompt injections are detected.
Gemini’s approach encourages combining these measures with sandboxing, human-in-the-loop verification, and strict access controls. Early enterprise adopters have already reported value from deploying these features in live environments.
Google continues to expand Gemini’s capabilities, positioning the platform to meet evolving demands for AI-powered automation in business settings. This latest update demonstrates Google’s focus on security, flexibility, and multi-environment support for enterprise customers.