Google integrates computer use natively into Gemini 3.5 Flash for agentic automation

Google has integrated computer use as a native built-in tool into Gemini 3.5 Flash, its main frontier agentic model. Previously, computer use was only available as a standalone Gemini 2.5 model; now it's part of the production 3.5 Flash release. The capability allows developers to build agents that can see, reason, and take action across browser, mobile, and desktop environments without leaving the main API.

This integration unlocks improved performance for long-horizon enterprise automation tasks like continuous software testing, knowledge work across professional applications, and multi-step orchestration. Early adopters have already demonstrated value: the model uses computer use to analyze its own features and audit documentation for accessibility issues. Developers can access the capability through the Gemini API and the Gemini Enterprise Agent Platform.

To mitigate prompt injection risks for agents operating in live environments, Google implemented targeted adversarial training. The release also includes two optional enterprise safeguard systems: one that requires explicit user confirmation for sensitive or irreversible actions, and another that automatically stops tasks if an indirect prompt injection is detected. Google recommends combining these with secure sandboxing, human-in-the-loop verification, and strict access controls.

For architects building production agents, this is a capability unlock: computer use is no longer a bolt-on experiment but a default tool in 3.5 Flash alongside function calling and grounding. Combined with 3.5 Flash's lower cost ($1.50/$9 per million tokens versus Claude Opus 4.7 or GPT-5.5) and faster execution (4x other frontier models), the integration removes friction from agentic RPA, workflow automation, and testing workflows.