If you're not using AI to attack your own systems, your adversaries will

Key Points:

• AI agents have demonstrated significant capabilities in hacking organizations, creating a new and complex attack surface that includes non-human identities and data-integration channels difficult to secure with traditional methods.
• Experts like Matt Hartman and Rob Joyce emphasize the growing need for organizations to adopt continuous, AI-enabled red teaming and penetration testing to keep pace with AI-driven attackers who operate relentlessly and at scale.
• Armadin, a startup founded by former Mandiant executives, conducted the largest controlled live AI cyberattack, showcasing how autonomous AI attacker swarms can simulate real-world threats far more efficiently than human teams.
• Traditional pen-testing practices, often conducted annually or quarterly, are no longer sufficient due to AI’s speed, scale, and sophistication; the cybersecurity industry is moving toward automated and continuous AI-driven testing while reskilling human pen-testers for broader roles.
• The evolving role of penetration testers now involves assessing business impact, testing AI system robustness, understanding agentic AI behavior, and developing harm taxonomies and guardrails to secure AI-integrated environments effectively.