Fastly Launches AI Runtime Control and Firewall for Enhanced Enterprise AI Security
Fastly introduces AI Runtime Control and AI Firewall to provide real-time visibility and governance over enterprise AI systems, addressing security and operational challenges as AI adoption accelerates.
Fastly has unveiled a suite of new capabilities designed to bring real-time control and security to enterprise AI deployments. The offerings, including AI Runtime Control and AI Firewall, aim to provide organizations with enhanced visibility and governance over their AI models, applications, and agents.
As businesses increasingly move AI from experimental phases into production, the need for robust security and operational management becomes paramount. Fastly's new tools are built upon its existing global edge platform, extending its proven capabilities in speed, security, and resiliency to the rapidly evolving AI landscape. The company notes a significant surge in machine-generated traffic, which now surpasses human traffic on its network, and a parallel increase in AI-related spending, with many organizations exceeding their AI budgets.
AI Runtime Control offers organizations granular control over AI model access and usage. It enables routing of model calls through a unified endpoint, supporting both public and self-hosted providers while using virtual keys to safeguard underlying credentials. This feature provides real-time insights into token expenditure, allows for rate limiting and budget enforcement, and facilitates failover mechanisms, ensuring model choice is preserved while central policy enforcement is maintained.
The AI Firewall is engineered to protect AI applications from emerging threats, particularly focusing on mitigating attacks like prompt injection. By analyzing prompts directly within the request path at the edge, the firewall can identify and block malicious inputs before they reach the AI models, thereby preventing unauthorized actions or data exfiltration.
Complementing these tools, Fastly's enhanced API Security capabilities address the growing challenge of AI agents interacting with enterprise APIs. These features enforce API contracts and apply consistent security rules to agentic, agent-assisted, and conventional traffic. This allows organizations to monitor or block non-compliant requests, preventing AI agents from accessing unauthorized operations or deviating from established API boundaries.
These new additions integrate with Fastly's existing portfolio, which includes services like Bot Management, ContentGuard, DDoS Protection, and Next-Gen WAF. By consolidating AI control and security onto the same infrastructure that already handles content delivery and application protection, Fastly aims to simplify the management of digital experiences and AI operations.
Kelly Shortridge, Chief Product Officer at Fastly, emphasized the importance of real-time production control for enterprises adopting AI agents and features. She stated that Fastly is extending the real-time control customers trust for content delivery and software security to the AI models, applications, and agents that are shaping the future of distributed systems. The goal is to enable organizations to innovate rapidly with AI while maintaining resilience and security.
These new offerings address critical areas of AI governance: how applications call models, how users and systems interact with AI applications, and how AI agents access enterprise APIs. As coding agents and AI-assisted development accelerate the creation of new endpoints, Fastly's platform aims to provide the necessary control and security to manage this expansion effectively.