F5 Labs Strengthens AI Security with Comprehensive Threat Intelligence and Monthly Leaderboards

F5 has launched enhanced threat intelligence resources aimed at helping enterprise security leaders assess and compare the risk profiles of various AI models. The F5 Labs threat research team has introduced the Comprehensive AI Security Index (CASI) and Agentic Resistance Score (ARS) leaderboards. These tools provide standardized benchmarks that are updated monthly, combining real-time attack intelligence with expert analysis of emerging AI attack vectors.

New Resources from F5 Labs

The introduction of these resources follows F5’s acquisition of CalypsoAI, which has significantly expanded its capabilities in AI security. The new offerings include one of the largest libraries of AI vulnerabilities, updated with over 10,000 new attack prompts each month and leveraging more than a year of accumulated attack data. F5 Labs aims to provide enterprises and the broader security community with a reliable method for evaluating, comparing, and selecting AI models and providers based on their ability to address critical challenges and real-world security threats.

Kunal Anand, Chief Product Officer at F5, emphasized the importance of verifying AI models before deployment in critical infrastructure. He stated that organizations require a continuous method to quantify resilience, and the F5 Labs AI Leaderboards serve this purpose by isolating specific weaknesses in the model layer. This intelligence enables security teams to govern inference and prevent attacks proactively.

Enhancing Security Visibility

The leaderboards from F5 Labs enhance visibility for security teams, allowing them to identify vulnerabilities comprehensively. This capability is designed to improve defenses in conjunction with the F5 Application Delivery and Security Platform, which protects APIs and data, mitigates DDoS attacks, and accelerates DevSecOps through automation and scalability.

The rapid integration of AI into various business operations necessitates robust security validation for the models organizations choose. F5 Labs’ leaderboards help security practitioners answer critical questions regarding model security by establishing metrics related to the paths of least resistance and the minimum computational resources required for both simple and complex attacks.

Key Metrics Offered by CASI

The Comprehensive AI Security Index (CASI) provides several important metrics, including:

• Average Performance: Baseline model performance measured across standardized tasks under normal operating conditions.
• Risk-to-Performance Ratio: Insight into the tradeoff between model safety and performance.
• Cost of Security: The current inference cost relative to the model’s CASI, assessing the financial impact of security.

Evaluating Resistance with ARS

In addition to CASI, the Agentic Resistance Score (ARS) evaluates how AI systems withstand sustained, adaptive attacks from AI agents with specific goals. Unlike traditional methods that focus on individual prompts, AI agents engage in prolonged interactions with models, employing reasoning and psychological tactics to bypass security measures. The ARS assesses AI systems across three core dimensions:

• Required Sophistication: The minimum level of attacker ingenuity necessary to compromise the AI system.
• Defensive Endurance: The duration for which the system remains secure under prolonged, adaptive, multi-step attacks.
• Counter-Intelligence: Whether failed attacks inadvertently reveal signals or system behaviors that could facilitate future exploits.

Advancements in AI Security

The recent announcement aligns with the general availability of F5 AI Guardrails and F5 AI Red Team. F5 AI Guardrails enable enterprises to implement both out-of-the-box and custom security measures to regulate how AI interacts with users and data. Meanwhile, F5 AI Red Team conducts rigorous testing using autonomous AI agents that simulate a dedicated team of security analysts, executing multi-step attacks that probe, learn, and adapt.

This comprehensive approach informs the CASI and ARS leaderboards, allowing F5 to measure both baseline model security characteristics and long-term resistance under real-world threat conditions. The CASI and ARS rankings are updated monthly, and F5 Labs also publishes monthly AI Security Insights, which provide detailed analyses of notable score movements, recent AI news, and specific AI attack deep-dives.

Grace Trinidad, Research Director for AI Security & Trust at IDC, noted that F5’s existing traffic inspection and logging capabilities have been enhanced through the acquisition of CalypsoAI. This expansion adds runtime AI governance capabilities and visibility into AI system behavior. The monthly updates to CASI and ARS rankings equip F5 customers with essential security metrics for selecting AI models and the tools necessary to manage how AI systems are utilized, what data they can access, and how outputs are generated.

F5 Labs’ AI Leaderboards exemplify the company’s commitment to helping customers and the security community navigate the complexities of AI security. By aggregating and presenting impactful results, F5 aims to better equip the global community to tackle the escalating challenges posed by modern cybersecurity threats. The formalization of a ranking system for leading AI models also promotes knowledge sharing among security professionals, fostering greater adoption of application security best practices.

For more information, visit securitymea.com.