Dual-use AI in cyberattack and cyber defence

Source article: AI-Powered Cyber Attacks and Defence Mechanisms

Artificial Intelligence (AI) is becoming an important component of modern cybersecurity. Machine Learning (ML), Deep Learning (DL), Natural Language Processing (NLP), and Generative AI are increasingly used to analyse security data, identify malicious behaviour, detect malware, classify phishing attempts, and support incident response. At the same time, the same technologies can be misused by cybercriminals to improve phishing, social engineering, synthetic-media fraud, vulnerability discovery, malware developme…

By the Truvace deskOctober 21, 20262 min read1 sourcePeer-reviewedTRV-2026-2128· cite & verify

Trace impact readingContested

P 69

Both sides are scored from claims and sources, not community votes.

G 68

In brief

This literature-based study reviews research and threat reporting from 2020 to 2026 on how artificial intelligence is used both to attack and to defend computer systems. It covers machine learning, deep learning, language processing and generative AI as tools for analysing security data and as tools for criminals.

The review matters because the dual-use nature of AI makes the security environment change quickly, so defenders cannot rely on AI alone. The paper says AI's usefulness depends on conditions such as data quality and oversight, and it treats privacy, ethics and adversarial machine learning as open concerns for future research.

Main points

Defensive uses include analysing security data, identifying malicious behaviour, detecting malware, classifying phishing and supporting incident response.
Offensive uses include better phishing, social engineering, synthetic-media fraud, vulnerability discovery and malware development.
AI effectiveness depends on data quality, continuous monitoring, secure model development, explainability and human oversight.
The paper recommends a layered architecture combining conventional security, AI analytics, continuous evaluation and human decision-making.

The gain

AI techniques can speed up and scale security analysis, helping detect malware, classify phishing and support incident response, provided data quality and human oversight are in place.

The problem

Cybercriminals can misuse the same AI technologies to improve phishing, social engineering, synthetic-media fraud, vulnerability discovery and malware development.

The rundown

The study is a literature review focused mainly on research and threat reporting from 2020 to 2026. It examines AI-assisted attack techniques, defensive applications, adversarial machine learning, and a comparison of traditional and AI-based security.

It also addresses real-world threat trends, limitations, privacy, ethics and future research directions, framing AI as a dual-use technology in cybersecurity.

The authors argue that AI should complement rather than replace established controls such as authentication, encryption, patching, network segmentation, endpoint protection and backups.

What this doesn’t fix

The paper is a literature-based review rather than an original experiment, and it states that AI defence effectiveness depends on data quality, monitoring, secure model development, explainability and human oversight.