SourceThe Evolving Role of the Red Team in the Era of Agentic Security
Autonomous AI agents allow cybercriminals to launch faster, larger-scale attacks with less human effort, reducing the time defenders have to react before an attacker reaches their goal.
- AI agents enable threat actors, particularly ransomware operators and access brokers, to massively scale attacks with reduced human effort and near-instantaneous dwell times.
- Open-weight models are projected to match frontier models' cybersecurity capabilities within 6 to 12 months, creating freely available tools for malicious actors.
- Malicious agent risks include industrialized social engineering, automated attack chains, cheapened supply chain attacks, low-cost offensive tooling, and rapid exploitation of hygiene issues.
- Google recommends Red Teams iteratively develop modular subagents connected via orchestrators to autonomously simulate agentic attacks and defend against emerging AI threats.