ARTIFICIAL INTELLIGENCE IN CYBERSECURITY: DUAL-USE TECHNOLOGIES, EMERGING THREATS, AND IMPLICATIONS FOR ECONOMIC SECURITY

ABSTRACT

Artificial Intelligence (AI) has become a cornerstone of modern cybersecurity, revolutionizing threat detection, anomaly identification, automated response mechanisms, and predictive analytics. At the same time, its dual-use nature enables malicious actors to develop more sophisticated, scalable, and evasive attacks, including AI-generated phishing campaigns, deepfake-enabled fraud, adversarial machine learning exploits, and automated ransomware operations. This article provides a comprehensive examination of these dynamics, analyzing the interplay between defensive innovations and offensive capabilities. The discussion extends to mitigation strategies encompassing the NIST AI Risk Management Framework, adversarial training, explainable AI (XAI), robust governance, public-private partnerships, and international cooperation. In an era of accelerating technological convergence and geopolitical competition, achieving an effective balance between AI-driven innovation and security is paramount for sustaining economic resilience and strategic autonomy.

Keywords: artificial intelligence, cybersecurity, dual-use technologies, economic security, deepfakes, adversarial AI, ransomware, AI governance, NIST AI RMF.

In today’s hyper-connected digital economy, artificial intelligence systems process vast quantities of data at unprecedented speeds, identify subtle patterns invisible to human analysts, and facilitate real-time decision-making that far surpasses traditional methods [1][7]. These attributes have positioned AI as a powerful ally in cybersecurity, enabling proactive defense through behavioral analytics for insider threat detection, predictive vulnerability assessment, malware classification, network traffic anomaly detection, and orchestrated incident response [2][8]. Organizations deploying mature AI-driven security solutions have reported substantial reductions in breach detection times and overall remediation costs [9].

Yet the very same technologies—generative models, large language models (LLMs), reinforcement learning, and computer vision—dramatically lower the barrier for cybercriminals, nation-state actors, and other adversaries. This allows them to automate attacks, personalize social engineering at scale, and evade traditional signature-based defenses [3; 10]. This dual-use dilemma creates unique and amplified risks to economic security, understood here as a nation’s capacity to maintain stable economic growth, protect critical digital infrastructure, safeguard sensitive data, ensure supply chain continuity, and preserve technological sovereignty in the face of external shocks and systemic vulnerabilities [4; 11].

Scholarly research and international policy reports consistently underscore AI’s role as a catalyst for a new arms race in cyberspace. On the defensive side, AI excels at processing massive datasets for threat intelligence and anomaly detection [5]. Conversely, offensive applications dramatically increase the volume, velocity, and sophistication of attacks. Generative AI can produce convincing phishing emails, voice clones, and deepfake videos in seconds, while adversarial techniques allow attackers to manipulate inputs to deceive security models or poison training datasets [6; 1].

The global COVID-19 pandemic and the ensuing digital acceleration further intensified these trends. Widespread remote work expanded attack surfaces, while the democratization of powerful AI tools made advanced capabilities accessible even to low-skilled actors. Studies document a sharp rise in AI-augmented incidents, with phishing campaigns achieving significantly higher success rates due to hyper-personalized, grammatically flawless content and real-time adaptability [2].

One of the most striking manifestations of AI’s offensive potential is deepfake technology in financial fraud [4]. Another critical domain involves adversarial machine learning. Attackers craft subtle perturbations to inputs—often imperceptible to humans—that cause AI security systems to misclassify malicious content as benign [5]. AI also supercharges ransomware and supply chain operations. Advanced persistent threat (APT) groups increasingly leverage generative models for reconnaissance, exploit development, and negotiation automation. Compromises targeting AI model repositories or poisoning publicly available training pipelines can propagate vulnerabilities across thousands of downstream applications [7; 10].

Geopolitical tensions add another layer of complexity. The concentration of advanced AI research, computing resources (including GPUs), and semiconductor manufacturing in a handful of countries creates strategic dependencies. Export controls, talent migration restrictions, and data localization policies heighten risks of deliberate disruptions. A successful AI-driven attack on critical infrastructure could cascade into widespread economic instability, including supply chain breakdowns, market volatility, energy shortages, and loss of public confidence [8; 11]. Emerging and developing economies often face heightened exposure due to limited resources for building resilient AI defenses, thereby exacerbating global digital inequalities [9].

The economic ramifications are extensive and multifaceted. Direct costs arise from data breaches, ransomware payments, and operational downtime. Indirect effects include elevated cybersecurity insurance premiums, increased compliance expenditures, deterred foreign investment due to perceived instability, and intellectual property theft that undermines long-term competitiveness. While AI adoption can reduce average breach costs, the overall expansion of the threat landscape often outpaces these gains, creating a persistent security-economics dilemma [2].

Addressing these challenges demands a holistic, multi-layered approach. The NIST Artificial Intelligence Risk Management Framework (AI RMF 1.0) offers a voluntary yet influential structure organized around the functions Govern, Map, Measure, and Manage. It emphasizes trustworthiness characteristics such as safety, security, resilience, transparency, and accountability throughout the AI lifecycle [1]. Organizations are encouraged to integrate these principles with existing cybersecurity frameworks.

Technical mitigations include adversarial training (exposing models to perturbed examples during development), robust data validation and provenance tracking, explainable AI (XAI) techniques to improve human oversight, and zero-trust architectures applied to AI pipelines. Continuous monitoring for model drift, regular red-teaming exercises, and secure model deployment practices are equally vital [5]. On the policy front, governments can provide incentives for domestic and allied AI R&D, mandate incident reporting for significant AI-related events, foster public-private threat intelligence sharing, and support the development of international norms and standards [10].

Beyond technology and policy, human factors remain crucial. Workforce training in AI literacy, ethical considerations, and responsible innovation helps bridge the gap between automated systems and human judgment. Ethical dilemmas—such as bias in security models, privacy implications of massive data collection, and the balance between open research and weaponization risks—require ongoing multidisciplinary dialogue [11].

In conclusion, artificial intelligence represents both the greatest opportunity and one of the most formidable challenges in contemporary cybersecurity. Its dual-use character necessitates proactive, collaborative, and adaptive strategies that integrate cutting-edge technology with sound governance and international cooperation. Nations and enterprises that successfully navigate this complex landscape—investing in resilient AI ecosystems, transparent development practices, and agile risk management—will be best positioned to protect their economic security and maintain strategic advantage in the 21st-century digital order. Failure to rise to this challenge risks ceding ground to adversaries and undermining the foundational trust upon which modern economies depend.

References:

1. Chlasta K. The dual-use dilemma of generative artificial intelligence in cybersecurity: Navigating the explosive growth in offensive and defensive applications // Security and Defence. – 2026.
2. Edris E.K.K. et al. Utilisation of Artificial Intelligence and Cybersecurity Capabilities: A Symbiotic Relationship // Electronics. – 2025. – Vol. 14, No. 10.
3. Jedrzejewski F.V. et al. Adversarial Machine Learning in Industry: A Systematic Literature Review // Computers & Security. – 2024.
4. NIST. Artificial Intelligence Risk Management Framework (AI RMF 1.0). National Institute of Standards and Technology, 2023. 48 p.
5. Pelekis S. et al. Adversarial machine learning: a review of methods, tools, and critical industry sectors // Artificial Intelligence Review. – 2025.
6. Rios Insua D. et al. Adversarial Machine Learning: Bayesian Perspectives // Journal of the American Statistical Association. – 2023.
7. Vassilev A. et al. Adversarial Machine Learning: A Taxonomy and Terminology of Attacks and Mitigations. NIST, 2025.
8. World Economic Forum. Artificial Intelligence and Cybersecurity: Balancing Risks and Rewards. 2025. URL: https://reports.weforum.org/docs/WEF_Artificial_Intelligence_and_Cybersecurity_Balancing_Risks_and_Rewards_2025.pdf(accessed: 16.05.2026).
9. UK Government. Cyber security risks to artificial intelligence. 2024. URL: https://www.gov.uk/government/publications/research-on-the-cyber-security-of-ai (accessed: 16.05.2026).
10. CoverLink. Cyber Case Study: $25 Million Deepfake Scam. 2025. URL: https://coverlink.com/case-study/case-study-25-million-deepfake-scam/ (accessed: 16.05.2026).
11. CSET. Adversarial Machine Learning and Cybersecurity. Georgetown University, 2025. URL: https://cset.georgetown.edu/publication/adversarial-machine-learning-and-cybersecurity/ (accessed: 16.05.2026).