Prompt injection and data theft risks in agentic LLM browsers are turning routine web bugs into full-session compromise paths, enabling attackers to hijack AI agents and exfiltrate sensitive data at scale silently. As agentic browsers like Perplexity Comet, OpenAI Atlas, Edge Copilot, and Brave Leo evolve from helpers into semi-autonomous operators, their hidden control planes […] The post Prompt Injection and Data Theft Risks Emerge In Agentic LLM Browsers appeared first on Cyber Security News.