Vijil Launched Automated Security Testing for AI Agents
The new Diamond Adaptive Red Teaming system uses adversarial agents to identify vulnerabilities in enterprise AI.
Updated on Oct. 7, 2026 in Artificial Intelligence
Live Poll
Do you trust that large companies are sufficiently securing their AI systems against security threats?
Vijil has released the Diamond Adaptive Red Teaming (DART) system to secure enterprise AI agents against evolving adversarial threats. The software utilizes automated multi-turn attacks to uncover security flaws and policy violations within agent frameworks.
Why it matters
As companies rapidly expand their use of autonomous AI agents, existing manual testing methods lack the speed and scalability to address sophisticated new attack tactics. DART provides a automated defense layer to keep pace with the growing complexity of these AI deployments.
DART employs multiple adversarial agents to conduct multi-turn attacks, which outperformed leading competitors in 9 out of 12 tested tasks. The system is designed to integrate directly into customer Virtual Private Cloud or on-premises environments.
The players
Vijil
Vijil is a San Francisco-based cybersecurity firm specializing in providing automated defense and testing solutions for enterprise-grade artificial intelligence agents.
The details
The system probes target defenses by adapting its tactics across various turns and episodes, ensuring it detects issues that static testing tools often miss. By integrating with existing deployment platforms, the software allows engineering teams to automatically test large fleets of agents without the need for manual oversight.
Timeline
October 07, 2026: Vijil released the DART system.
2025: Fortune 500 companies utilized an average of fewer than 15 agents.
2028: Projections indicate Fortune 500 companies will use over 150,000 agents.
The Tech Race
The transition from human-managed systems to autonomous AI agents requires a fundamental shift in how corporations approach cybersecurity. This development marks a move toward machine-speed defense, replacing manual auditing with continuous, adversarial machine-versus-machine testing.
For enterprise users and developers, this integration allows for the continuous monitoring of agent fleets without added infrastructure costs. It minimizes the manual workload for engineering teams while providing more robust protection against unauthorized agent behavior.
The takeaway
Automated adversarial testing is becoming a necessary component of enterprise AI as the volume of agents in production continues to scale. Organizations should focus on integrating these autonomous security layers early in the development lifecycle to mitigate the risks of evolving attack tactics.
Further reading
For broader context on the evolving security landscape for automated systems, visit our Artificial Intelligence section.
Live Poll
Do you trust that large companies are sufficiently securing their AI systems against security threats?
