Google's Chrome team built a separate, closed online environment it calls the Mirror Dimension, an alternate internet it fully controls. It lets engineers train and test Gemini-powered AI browsing agents without exposing the real web to unpredictable behavior. The project was reported by WIRED.
Google's Chrome team has built an entire alternate internet to keep its AI browsing agents in check. The company calls it a sandbox, and it is one the team completely controls.
What Is Google's Mirror Dimension Sandbox?
The effort grew out of work on AI browsing features powered by Google's Gemini models. As the team developed those features, it realized it needed extra safeguards to keep the agents from going haywire. The solution was a separate online environment, which the company describes as an expansive, alternate-universe internet.
Why Do Chrome AI Agents Need a Closed Test Environment?
That environment gives the Chrome team room to train AI browsing agents without exposing the real web to the risks of an agent behaving unpredictably. Because the internet is closed off and under Google's control, engineers can test how the agents navigate and act without the consequences that would follow on the open web.
The project was described in a WIRED report by Lily Hay Newman, a senior writer who covers information security, digital privacy and hacking. The story was published Nov. 19.
What Risks Do AI Browsing Agents Pose on the Open Web?
The Mirror Dimension approach highlights a practical problem facing companies that build AI agents meant to carry out tasks in a web browser. An agent that clicks, types and submits forms on a person's behalf can cause real damage if it misfires. Building a controlled test environment is one way to find those failures before the features reach users.
Dieser Artikel wurde mit Unterstützung von KI verfasst.
News Factory APP - agentische News für besseres SEO & AEO.
