Gemini 2.5 Computer Use gives agents direct browser control
Google released Gemini 2.5 Computer Use in public API preview for agents that click, type, scroll, and operate behind logins. Google also documented per-step safety checks, developer-defined confirmations, and restrictions for high-risk actions, while confirming production use in UI testing and agentic Search features.
0 comments · 0 votes
Sign in to join the discussion →
No comments yet. Start the discussion.
Why it moved the index
A broadly accessible model that directly operates graphical interfaces, including logged-in workflows, materially expands consequential agent autonomy beyond structured APIs. Primary launch and model-card evidence establish access, production use, and explicit safeguards. Magnitude is moderated because it was a preview and actions were executed by a client loop with safety services and confirmation controls.
Assessment history
-
R1
Toward 66 · confidence 93
New dated primary-source launch and model-card evidence absent from the durable catalogue.
14 Aug 2026
Share this page
-
DoomBench assesses “Gemini 2.5 Computer Use gives agents direct browser control” as evidence moving toward doom, with magnitude 66 and confidence 93 out of 100 in the autonomy and agency category.
-
The DoomBench assessment of “Gemini 2.5 Computer Use gives agents direct browser control” is based on reporting from Google and records the editorial rationale, source quality, attribution, and revision history.
-
DoomBench summarizes “Gemini 2.5 Computer Use gives agents direct browser control” as follows: Google released Gemini 2.5 Computer Use in public API preview for agents that click, type, scroll, and operate behind logins. Google also...
https://www.doombench.com/news/gemini-2-5-computer-use-gives-agents-direct-browser-control-2025-10-07