Spy vs. spy: How GenAI is powering defenders and attackers

This report surveys the evolving role of generative and agentic AI in cybersecurity, outlining how adversaries leverage AI for coding, phishing, evasion (including prompt-based anti-analysis in malware and DNS), and vulnerability discovery, while defenders apply it to threat intel processing, incident response, and proactive code analysis. It highlights the growing accessibility of uncensored/open-weight models, the potential surge in adversarial capability as models shrink, and the supporting role of technologies like MCP servers. The piece emphasizes that current AI-driven attacks still require significant human involvement and that expertise remains critical to effectively harness AI on both sides.