What question did this study set out to answer?

The research evaluates the effectiveness of large language models in cybersecurity tasks based on existing evidence.

April 10, 2026Open Access

Artificial Intelligence in Cybersecurity: Reproducible Evidence on Vulnerability Discovery and Security Operations

SPSofia PlajutinBose (United States)JRJulian Ignacio RamirezBose (United States)

Key Points

The research evaluates the effectiveness of large language models in cybersecurity tasks based on existing evidence.
Synthesize publicly available evidence on LLMs in cybersecurity
Audit Mozilla advisories and peer-reviewed security studies
Assess regulatory standards for LLM deployment in cybersecurity tasks
LLMs are effective in vulnerability analysis and penetration testing
LLMs can augment incident response processes
Claims of fully autonomous security operations lack support without governance controls

Abstract

This paper synthesizes publicly available evidence on the role of large language models (LLMs) in cybersecurity operations. Rather than proposing new benchmarks, the work audits existing research—including Mozilla advisories, peer-reviewed security studies, and regulatory standards—to assess which cybersecurity tasks are currently well-supported, weakly supported, or not supported for LLM deployment. The authors argue that current evidence supports LLM-assisted vulnerability analysis, penetration-testing workflows, and incident-response augmentation, but does not support claims of fully autonomous security operations without substantial governance controls. The paper contributes a machine-readable evidence matrix and task-level summary to enable independent verification of its claims.

AI से पूछें

Bookmark

View Full Paper