Real World Impact of Generative AI (GenAI) on Cybersecurity Professionals

Sammanfattning

Background: Rapidly evolving cyber threats challenge conventional cybersecurity measures, positioning Generative Artificial Intelligence (GenAI) as a transformative force. GenAI offers potent defensive capabilities such as enhanced threat detection and automated responses, but also introduces significant offensive risks, including sophisticated attacks. This duality creates opportunities and challenges, highlighting a discrepancy between GenAI’s theoretical potential and its practical integration in cybersecurity. Purpose: This thesis investigates the adoption of GenAI by cybersecurity professionals, focusing specifically on the discrepancy between its widely theorized potential and its actual application in real-world operational settings. It analyzes the gap between the anticipated benefits of GenAI and the reported experiences of professionals utilizing it. To this end, the study explores key dimensions influencing GenAI adoption, such as its perceived usefulness and ease of use (PU & PEOU), the alignment of its capabilities with cybersecurity task requirements (Task-Technology Fit/TTF), influential organizational and technological factors, and the practical challenges encountered during adoption. The central research question seeks to determine the nature and extent of this theory-practice gap. Ultimately, the study provides a nuanced understanding of GenAI’s practical value and limitations in cybersecurity. Method: This study employed a qualitative methodology to explore GenAI adoption experiences. Ten cybersecurity professionals, purposively selected for their hands-on experience with GenAI in their roles, were interviewed semi-structured. The collected data was analyzed using thematic analysis, with findings interpreted through an integrated theoretical framework combining the TAM and TTF theories. Results and analysis: The analysis, framed by the integrated TAM-TTF model, revealed that cybersecurity professionals perceive GenAI as highly useful (PU) for specific tasks, where a strong TTF enhances efficiency. However, its broader adoption is significantly hampered by several critical factors. Notably, achieving proficient PEOU requires substantial learning and adaptation effort. More critically, pervasive issues with GenAI’s reliability and explainability severely undermine user trust, a crucial element for critical cybersecurity functions. Furthermore, substantial perceived risks related to data privacy, security vulnerabilities, and output integrity contribute to cautious adoption patterns. Consequently, GenAI is predominantly viewed as a powerful assistive tool that necessitates considerable human oversight and validation, with the gap between its anticipated benefits and actual experiences defined mainly by these trust, usability, and risk concerns, especially for high-stakes operations. Conclusions: The study concludes that a significant gap exists between GenAI’s anticipated benefits and professionals’ actual experiences. This discrepancy is primarily driven by current limitations in GenAI’s reliability and explainability, compounded by challenges and perceived risks. Thus, GenAI currently operates as an assistive tool requiring considerable human oversight, not an autonomous solution. The extent of this theory-practice gap varies, being most pronounced for complex functions demanding high reliability and transparency. Recommendations for future research: Future research directions include more granular evaluations of TTF for specific cybersecurity tasks, effective XAI to build practitioner trust, GenAI skills development, and long-term research observing how GenAI adoption and its impact shift as the technology itself improves.