AI is driving a growing share of automated traffic and AI traffic is increasing 6.5x faster than human traffic. In July 2026, machine generated traffic became the majority of traffic across Fastly's network for the first time.
These shifts impact every decision maker responsible for enterprise infrastructure.
AI usage is causing three challenges for infrastructure leaders: it creates blind spots, complicates cost and performance controls, and introduces new vulnerabilities that are hard to secure.
To tackle these issues, we’ve expanded our Fastly for AI portfolio—a suite of tools designed to govern, protect, and control your entire AI stack on the fast, secure global platform you already rely on.
Govern Your AI Usage with AI Runtime Control
Your engineers are building with AI, embedding models into internal tools and spinning up agents across the business. Nobody can tell you exactly who is using which model and what prompts they are sending.
Fastly’s AI Runtime Control removes the blind spots. It captures every single AI call and LLM request your business makes. Calls are mapped to virtual keys and specific client sessions. You can set hard budget limits, enforce rate limits per key, and automatically fail over between providers if one goes down.
Take Back Control of Your AI Spend
Most engineering and IT teams have a massive shadow AI problem. McKinsey found that 93% of organizations are exceeding their AI budgets, and companies that are strategic with their AI consumption can save 20 to 30% of the overall AI costs.
AI Runtime Control helps you manage your AI spend and gives you back the control. You get:
Full visibility: Searchable prompt and completion logs to see exactly what your AI is doing.
Real-time cost control: Track token usage and exact costs across every major model provider plus self-hosted and custom models.
Low-latency performance: Everything is processed at the edge, eliminating the round-trip delays of cloud-hosted gateways.
The setup requires no code changes and plugs directly into existing observability tools. Read our step-by-step AI Runtime Control setup guide.
Protect Your AI Applications with AI Firewall
Every customer-facing AI chatbot creates a new attack surface. Attackers hide malicious instructions inside ordinary web pages or support tickets, and when your AI reads that content, the agent gets hijacked.
Legacy firewalls rely on patterns, while Fastly’s Next-Gen WAF provides modern application and API defense against standard web attacks. Adding AI Firewall extends that protection to your AI models. It acts as a checkpoint for your AI and inspects incoming requests for malicious commands before they can do harm:
Proactive blocking: It scans traffic at the edge, catching and blocking malicious instructions before they reach your AI model.
High performance: It runs on standard hardware, meaning it requires no special GPU power.
Layered security: It serves as your first line of defense. It is designed to catch common threats and works best alongside your existing security tools, rather than replacing them.
Pro tip: For comprehensive security, we recommend using AI Firewall with Fastly’s core security suite. Fastly Next-Gen WAF, Bot Management with ContentGuard, Fastly DDoS Protection, and API Security safeguard your applications, APIs, and infrastructure from advanced threats. When you pair them with AI Firewall you get purpose-built protection aligned with the latest AI threat frameworks.
Control Your AI-Powered APIs at Scale
Larger enterprises with over $10 billion in annual revenue manage an average of 1,400 APIs. When agents make unpredictable or broken API calls, your servers pay the price. The best way to prevent attacks is to secure the traffic right where it flows—at the edge.
Fastly’s new API Security capability - API Security schema enforcement - acts as a runtime control. By uploading your API schema, you can validate every live request against it. If an agent tries to execute a call that breaks your published contract, you can block it instantly in blocking mode, or log it first if you want to observe before enforcing.
API Security schema enforcement keeps traffic conformant with the API contract you already published. It brings security to APIs never designed for AI, as well as the specialized servers that agents rely on to function. And it is not just for AI apps! You can turn it on today for any traffic.
Ready to govern, protect and control your AI operations? AI Runtime Control, AI Firewall, and API Security provide the layered control you need. Explore Fastly for AI.