Introducing computer use in Gemini 3.5 Flash
Key Points
- Built-in computer use in Gemini 3.5 Flash
- Adversarial training plus optional enterprise safeguards
- Accessible via Gemini API and Enterprise Agent Platform
Summary
Gemini 3.5 Flash now includes computer use as a built-in capability, enabling developers to build agentic systems that see, reason, and take actions across browser, mobile, and desktop environments. The integration brings improved performance for long‑horizon and enterprise automation tasks (for example, continuous software testing and knowledge work across professional applications) and complements existing function-calling and tool-grounding features.
Key Points
- Built-in computer use: native integration in Gemini 3.5 Flash (previously available only as a standalone Gemini 2.5 computer use model).
- Primary engineering use cases: automated cross-application workflows, continuous UI/software testing, enterprise automation and knowledge work.
- Safety measures:
- Targeted adversarial training to mitigate prompt-injection risks.
- Two optional enterprise safeguards: require explicit user confirmation for sensitive/irreversible actions; automatically stop tasks when indirect prompt injection is detected.
- Recommended defense-in-depth: secure sandboxing, human-in-the-loop verification, strict access controls.
- Getting started: evaluate capabilities in the Browserbase demo, then build using the Gemini API or the Gemini Enterprise Agent Platform with the provided reference implementation and documentation.