F5, a provider of application and API delivery and security solutions, has introduced threat intelligence tools intended to help enterprise security teams assess and compare the risks associated with AI models. Using updates from F5 Labs, organisations gain monthly insights from the Comprehensive AI Security Index (CASI) and the Agentic Resistance Score (ARS) leaderboards.
These resources follow F5's acquisition of CalypsoAI, which adds an AI vulnerability library, updated with more than 10,000 new attack prompts each month. The tools aim to assist enterprises in evaluating AI models prior to deployment, considering real-world application security threats.
F5 Labs’ leaderboards provide metrics including:
- Average Performance: Baseline model performance across standardised tasks under normal conditions.
- Risk-to-Performance Ratio: The relationship between model safety and efficiency.
- Cost of Security: The financial impact of implementing security measures relative to the model’s CASI score.
The ARS assesses how AI systems withstand sustained, adaptive attacks, examining required sophistication, defensive endurance, and counter-intelligence capabilities.
These resources aim to complement F5’s existing tools, such as AI Guardrails and AI Red Team, which focus on securing AI interactions with users and testing models using autonomous AI agents.
By integrating F5’s security platform with CalypsoAI’s capabilities, the CASI and ARS metrics provide enterprises with insights to evaluate and monitor AI model security. The initiative also provides reference points for the broader security community, aiming to support the adoption of application security best practices and AI governance in enterprise environments.
