AITechTalks5 months ago
Perplexity BrowseSafe fails 36% of prompt injection tests
1 min read
Lasso Security achieved a 36% bypass rate against Perplexity's BrowseSafe using standard encoding techniques. The open-source tool designed to protect AI browsers from prompt injection failed against attacks wrapped in HTML structures. The findings demonstrate that single-model security creates dangerous blind spots for agentic browsers.
Level
Hype check
Tap to vote and see what everyone thinks.
In this storyPerplexity
