Google builds fake internet to hack its own Chrome AI agents
Google's Chrome AI Security team detailed "Mirror Dimension," a sandboxed synthetic internet of fake social, shopping and banking sites used to test its browsing agents. Google says each security engineer's ideas become more than 1,000 attack test candidates a day, over 30 times the manual rate. A three-role agent harness builds the sites: a research agent writes specs, a builder agent writes the code, and critic agents check the results. The report notes there are no figures on how often Chrome's agent resists the attacks.
From aiweekly.co's report.
Why it matters: Google is stress-testing its Chrome AI browsing agents against fake banking and shopping sites before release, which affects everyday users, though with no published figures on how well the agent resists attacks, its safety remains unproven.
Comments
No comments yet.