Skip to content
Search
paper2026Unreviewed

Securing LLM Powered AI Browsers Against Prompt Injection: A Comprehensive Survey, Threat Taxonomy, and Defense Framework

Sabin Adhikari, Roshan Paudel, Dipesh Gautam, Sanjog Chhetri Sapkota, Biplab Dahal, Roshit Raj Paudel, Anupam Dhakal

View resource

Abstract

Prompt injection is a serious threat to the security of large language models operating in AI-powered browsers and autonomous web agents, which depend on the ability of those models to interpret instructions correctly as they are used for automated browsing, data extraction or content processing. The inherent shortcomings of these systems to confidently differentiate trusted instructions from malicious ones allow the threat to influence agent behavior, exfiltrate sensitive information, and circu

Categories

Prompt InjectionMembership Inference

Framework mappings

OWASP Top 10 for LLM Applications
  • LLM01Prompt Injection
  • LLM02Sensitive Information Disclosure
MITRE ATLAS
  • AML.T0024.000Infer Training Data Membership
  • AML.T0051LLM Prompt Injection

Suggested from the entry's categories.

Cite

@misc{adhikari2026securing,
  title = {{Securing LLM Powered AI Browsers Against Prompt Injection: A Comprehensive Survey, Threat Taxonomy, and Defense Framework}},
  author = {Sabin Adhikari and Roshan Paudel and Dipesh Gautam and Sanjog Chhetri Sapkota and Biplab Dahal and Roshit Raj Paudel and Anupam Dhakal},
  year = {2026},
  doi = {10.2139/ssrn.6340078},
  url = {https://doi.org/10.2139/ssrn.6340078}
}