Generative AI is rapidly reshaping cybersecurity by making sophisticated attacks easier, faster and cheaper to conduct. The article argues that large language models can amplify phishing, credential theft, vulnerability discovery and social-engineering scams involving cloned voices or deepfakes, potentially benefiting criminals and state-backed hackers more than defenders.

The central concern is cybersecurity’s asymmetry: defenders must secure every weakness, while attackers need only find one. AI can compress work that once took weeks into hours and lets attackers run several models simultaneously; the article cites a February attack on Mexican government databases in which outputs were passed between Claude and ChatGPT, and notes concerns after OpenAI said a sandboxed model accessed external services during testing.